Integrated multi-objective management of saltwater intrusion in coastal aquifers using coupled simulation-optimisation and monitoring feedback information

Janardhanan, Sreekanth (2012) Integrated multi-objective management of saltwater intrusion in coastal aquifers using coupled simulation-optimisation and monitoring feedback information. PhD thesis, James Cook University.

[img]
Preview
PDF (Thesis)
Download (1MB) | Preview
 
240


Abstract

Coastal aquifers are aquifers which are hydraulically connected to the sea. They are important sources of groundwater which are often over-exploited due to the high density of population existing near the coasts. Coastal aquifers are susceptible to seawater intrusion caused by over-exploitation or other factors like sea level rise due to climate change. Carefully planned groundwater extraction and monitoring strategies are required for the optimal and sustainable use of coastal aquifers. This study develops methodologies for multi-objective optimal groundwater extraction strategies using simulation and optimisation techniques. Two conflicting objectives of management, viz, maximising the total beneficial pumping and minimising the total barrier well pumping are considered in this work. This study also develops optimal monitoring network designs for evaluating the compliance of the implemented strategies with the prescribed ones and illustrates the sequential modification of the prescribed strategies based on the feedback information from the compliance monitoring network.

A coupled simulation-optimisation framework is proposed and developed as the basic tool for deriving optimal groundwater management strategies. A three dimensional density-dependent flow and transport simulation model FEMWATER is used to simulate the coastal aquifer responses to groundwater extraction, in terms of the saltwater intrusion levels. A large number of such simulations is performed to generate the concentration levels resulting from different combinations of pumping from the beneficial and barrier well pumping locations. This pumping-salinity dataset is used as input-output patterns to train and test surrogate models based on modular neural networks (MNN) and genetic programming (GP). Properly trained and tested surrogate models are coupled to the multi-objective genetic algorithm. The optimisation algorithm iteratively searches for the optimal groundwater extraction strategies in a number of generations and in each step, the surrogate models are run to evaluate the salinity levels resulting from the candidate pumping strategies considered. The Pareto-optimal set of solutions is evolved after a number of such generations. It is observed from the obtained results that both surrogate modelling approaches identified similar Pareto-optimal front of solutions for the coastal aquifer management problem. However, the genetic programming based surrogate modelling approach is found to have specific advantages when used in the simulation-optimisation framework.

One of the main concerns regarding surrogate modelling based simulation-optimisation is the non-reliability issues associated with the optimal solutions resulting from the approximation involved and predictive uncertainty of the surrogate models. In this study a methodology is developed for obtaining reliable solutions to coastal aquifer management by overcoming the predictive uncertainty of the surrogate models. In this approach, an ensemble of surrogate models is developed to predict the aquifer responses to pumping. Bootstrap samples of pumping-salinity patterns are used to train and test different surrogate models using genetic programming. The number of surrogate models in the ensemble is determined by an uncertainty criterion. All the surrogate models in the ensemble are independently coupled to the multi-objective genetic algorithm, and a multiple-realisation optimisation approach is utilised to derive reliable optimal pumping strategies for coastal aquifer management. Reliability of optimal solutions is defined in terms of the percentage of the surrogate models, for which the imposed constraints are satisfied in deriving the pumping solutions. From these results, it is observed that optimal solutions with increased levels of reliability can be obtained using this proposed approach. The ensemble surrogate based methodology is further extended to address coastal aquifer management under parameter uncertainty. Uncertainty in the values of hydraulic conductivity and annual aquifer recharge are considered. The realisations of hydraulic conductivity and aquifer recharge are sampled from their respective distributions using Latin hypercube sampling. Bootstrap samples of pumping-salinity patterns generated using the numerical simulation model over different realisations of the uncertain parameters are used to train and test different surrogate models in the ensemble. Thus, surrogate models in the ensemble have different predictive capabilities in different regions of the parameter-decision space. All the surrogate models are then coupled with the multi-objective genetic algorithm, and multiple-realisation optimisation is performed incorporating the reliability criterion. This approach results in the robust optimisation of the groundwater management strategies under parameter uncertainty. On validating the derived optimal solutions with the numerical simulation model for different realisations of the uncertain parameters, it was observed that these solutions are robust for the range of values of the uncertain parameters considered. For performance evaluation, the methodology is applied to an existing well field in a realistic coastal aquifer system in the Lower Burdekin in Australia.

Compliance monitoring is an essential component of any groundwater management project. A methodology is developed for the design of compliance monitoring networks in this study. The network is necessary to monitor the compliance of the actual field level implementation with the simulated results. The design is performed subject to the imposed constraint of budgetary limitation, implemented as the maximum permissible number of monitoring wells. Subject to this constraint, the design methodology incorporates two goals within a single objective, viz, to place the monitoring wells where there is maximum uncertainty and to reduce the redundancy in monitored information by minimising the coefficient of correlation between the monitored locations. This objective of monitoring network design is compared against the widely used objective of uncertainty maximisation and the advantages are illustrated. The use of compliance monitoring information to sequentially update the coastal aquifer management strategies is illustrated by simulation experiments. A deviation from the prescribed optimal strategy, at any stage during the field implementation, may result in undesirable effects like increased levels of salinity. Based on the compliance monitoring information, the pumping strategies for the subsequent stages of management are modified to compensate for these ill effects. The results of the simulation experiments conducted for the Lower Burdekin aquifer illustrate that sequential updating of the management strategies based on compliance information helps to better achieve the objectives of management.

A coupled simulation-optimisation framework using trained and tested surrogate models based on genetic programming are shown to be computationally efficient tools for developing optimal extraction strategies for coastal aquifer management. The newly developed ensemble surrogate modelling with multiple realisation optimisation has potential applications in deriving reliable and robust strategies for coastal aquifer management under parameter uncertainty. The developed simulation-optimisation methodology for developing optimal pumping strategies, together with the designed compliance monitoring network and sequential updating of the strategies constitute an integrated approach for the management and monitoring of coastal aquifer systems.

Item ID: 36942
Item Type: Thesis (PhD)
Keywords: aquifers; coast; coastal; design; engineering; flow; ground water; groundwater; intrusion; management; modeling; modelling; models; optimal; optimizing; salt water; saltwater; simulation; simulation-optimisation; simulation-optimization
Related URLs:
Additional Information:

Publications arising from this thesis are available from the Related URLs field. The publications are:

Sreekanth, J., and Datta, Bithin (2010) Multi-objective management of saltwater intrusion in coastal aquifers using genetic programming and modular neural network based surrogate models. Journal of Hydrology, 393 (3-4). pp. 245-256.

Sreekanth, J., and Datta, Bithin (2011) Coupled simulation-optimization model for coastal aquifer management using genetic programming-based ensemble surrogate models and multiple-realization optimization. Water Resources Research, 47. pp. 1-17.

Sreekanth, J., and Datta, Bithin (2011) Comparative evaluation of genetic programming and neural network as potential surrogate models for coastal aquifer management. Water Resource Management, 25 (13). pp. 3201-3218.

Sreekanth, J., and Datta, Bithin (2011) Optimal combined operation of production and barrier wells for the control of saltwater intrusion in coastal groundwater well fields. Desalination and Water Treatment, 32 (1-3). pp. 72-78.

Sreekanth, J., and Datta, Bithin (2012) Comment on "Artificial neural network model as a potential alternative for groundwater salinity forecasting" by Pallavi Banerjee et al. [J. Hydrol. 398 (2011) 212–220]. Journal of Hydrology, 420-421. pp. 419-420.

Sreekanth, J., and Datta, Bithin (2014) Stochastic and robust multi-objective optimal management of pumping from coastal aquifers under parameter uncertainty. Water Resources Management, 28 (7). pp. 2005-2019.

Sreekanth, J., and Datta, Bithin (2011) Wavelet and cross-wavelet analysis of groundwater quality signals of saltwater intruded coastal aquifers. In: Proceedings of the 2011 World Environmental and Water Resources Congress (414), pp. 846-853. From: World Environmental and Water Resources Congress 2011: Bearing Knowledge for Sustainability, 22-26 May 2011, Palm Springs, CA, USA.

Date Deposited: 22 Dec 2014 23:25
FoR Codes: 09 ENGINEERING > 0905 Civil Engineering > 090509 Water Resources Engineering @ 100%
SEO Codes: 96 ENVIRONMENT > 9609 Land and Water Management > 960903 Coastal and Estuarine Water Management @ 50%
96 ENVIRONMENT > 9609 Land and Water Management > 960999 Land and Water Management of Environments not elsewhere classified @ 50%
Downloads: Total: 240
Last 12 Months: 1
More Statistics

Actions (Repository Staff Only)

Item Control Page Item Control Page