Determination of Maximum Groundwater Level Surface

Tara Zirakbash

Urban development impacts groundwater levels, and in turn, groundwater variations may result in major risks to infrastructure and the environment and cause economic losses. Uncertainty in determining maximum groundwater levels poses significant challenges in the design of drainage, road construction, and foundations of structures and causes difficulties for building developers. Failure to identify appropriate separation distance between groundwater levels and urban structures could cause damage and risks such as inundation of road surfaces, drying surface water bodies, reduced efficiency of stormwater infiltration systems and quality issues. Currently, there is no standard methodology for determining the maximum groundwater level surface, and efforts to develop reliable methods for its optimal spatial prediction have been minimal. This project aimed to develop a fit-for-purpose methodology for determining the maximum groundwater level surface through numerical modelling and model-independent parameter estimation (PEST). This thesis makes its contribution in the following steps. This research first evaluated three conventional methods used to calculate the maximum groundwater level in Western Australia with similar principles. The goal was to identify the reasons and range of differences in results by applying the methods to assess the reliability of their calculations at monitoring locations and other areas across a study area in Perth. Shortcomings related to changes over time by human interventions and climate, various sample data densities, the inconsistent period of data selection and lack of consideration of physical concepts in the application of those methods were identified. It was determined that the Design Groundwater Level Method (DGWL), which provided the data for this research, was the most efficient, and the quality of its data at the monitoring locations was sufficiently accurate for the purpose of this research. This research aimed to assess the enhancement of spatial prediction accuracy by using a numerical flow model as an interpolation tool. A common approach to understanding the spatial variability of groundwater levels is using various interpolation techniques. Therefore, this study investigated the comparative accuracy of ten interpolation methods implemented in ArcGIS to provide insights on choosing types of interpolation methods. Also, it elucidated that the narrow spatial distribution of the monitoring network could increase the accuracy of prediction and characterised the locations of large prediction errors related to changes in physical characteristics. The outcome of this section was used as a basis for comparison with the results of the numerical model. Then the spatial prediction accuracy of the groundwater level surface was evaluated by using a hybrid approach of data-driven and physical models. It utilised the maximum groundwater levels at the monitoring locations estimated by DGWL using statistical analysis and utilised a steady-state numerical groundwater flow model for spatial prediction. For the performance evaluation of the model, the cross-validation technique, which is widely used in the field of machine learning, was utilised. The findings indicated that automatically calibrated numerical models surpassed interpolation methods. The numerical model had twice the spatial accuracy of interpolation for one dataset and comparable performance for another. The numerical model's results were less dependent on dataset specifications compared to interpolation. Finally, the risk of overfitting was evaluated, and several regularisation settings were tested to gain insight into the optimal regularisation settings and surface prediction. Moreover, the linear predictive uncertainty and error variance were examined to determine the usefulness of calibration and the uncertainty level that remained in the results. Furthermore, a practical solution by using the cross-validation approach was proposed as a means of preventing overfitting. The resultant accuracy measures ascertained that a simplified groundwater numerical flow model has levers to consider the nuance of the groundwater flow and potentially improve the accuracy of prediction and prevent overfitting compared with interpolation techniques. This research project can inform water resource planning, design, and management for urban development while identifying future research directions for maximum groundwater levels. Accurate determination of the maximum groundwater level is essential for preventing risks, sustainable development and cost-effective design. The results can guide hydrogeologists, engineers, and management authorities in assessing groundwater levels and developing effective strategies.

Determination of Maximum Groundwater Level Surface

Files and links (1)

Abstract

Details

Metrics