Quantification of the probable environmental effects of the Hinds Managed Aquifer Recharge trial using mathematical modelling and advanced uncertainty techniques.

Abstract

Internationally, Managed Aquifer Recharge (MAR) has gained recognition as a mechanism to address environmental degradation. Investigations into its effectiveness in the New Zealand setting are ongoing and started with a five-year trial near Hinds, mid-Canterbury. The Hinds MAR trial aimed to raise groundwater levels, improve lowland streamflow and improve both ground and surface water quality. This research used advanced numerical modelling techniques such as null space Monte Carlo to assess the probable effectiveness of the Hinds MAR trial. Use of numerical uncertainty analysis to understand the probable effects of MAR programmes is both recommended and noted as a gap in international literature. Secondly, this research investigates the usefulness of various modelling approaches for quantifying the effects of MAR on the receiving environment.

The water resources of the Hinds Plains, mid-Canterbury, have been degraded by decades of high- intensity agriculture. The effects are seen as lowered groundwater levels, reduced spring-fed stream flows and high nitrate-nitrogen concentrations in both surface and groundwater. The Hinds MAR trial sought to introduce 500 l/s recharge into the groundwater system. The original evaluations of the trials probable effectiveness are based on this rate of recharge. However, due to in-situ conditions, the design recharge rate was never realised. Using the model code MODFLOW-NWT and advanced uncertainty analysis, this research investigated the probable response to the long-term achievable recharge rate of approximately 110 l/s after the trial completion.

This research investigated the effectiveness of a range of modelling approaches, from simple analytical and homogenous parameter numerical models to highly parameterised, spatially variable numerical models. It ultimately settles on highly parameterised numerical modelling as the most effective approach to assess the effects of the MAR trial. Like previous international studies, this research demonstrates the importance of calibrating numerical models to both quantity and quality, especially if useful results are to be obtained for water quality outcomes. For instance, initial uncertainty analysis used a model that was only calibrated to quantity, and despite a suite of 100 calibration constrained simulations it was not possible to reproduce observed water quality changes in response to the trial. Failure to capture the hydraulic conductivity field that represented the observed water quality changes was possibly caused by the model calibration being biased towards a local calibration optima, something highlighted in international literature as a possible unwanted outcome with calibration constrained uncertainty analysis. Recalibration of the flow model with the inclusion of concentration targets as calibration criteria produced more reasonable approximations of the observed and expected water quality changes.

The final suite of null space Monte-Carlo simulations suggest the Hinds MAR trial will successfully raise groundwater levels across a large area and increase stream flows. Further, the trial will improve water quality in groundwater, though it will probably not influence surface water quality. Transport modelling suggests water quality improvements can be expected for several kilometres down-gradient of the trial site, though they are unlikely to propagate as far as the lowland streams.

In terms of the appropriateness of the various modelling techniques investigated, analytical modelling is likely sufficient to estimate the mounding effects in the immediate vicinity of the trial. However, once the area of interest extends beyond the immediate trial site, numerical modelling should be applied. Water quality observations should be included in the calibration targets, and flow and transport calibrated simultaneously. Uncertainty analysis was useful for providing confidence in modelled outcomes and should be employed if the risks (whether financial or environmental) associated with MAR programmes need to be considered.