Elsevier

Journal of Hydrology

Volume 588, September 2020, 125091
Journal of Hydrology

Research papers
Evaluating robustness of waste load allocation under climate change using multi-objective decision making

https://doi.org/10.1016/j.jhydrol.2020.125091Get rights and content

Highlights

  • A model framework is advanced for recognizing robust waste load allocation (WLA) plans under climate change.

  • Coupled SWAT and BorgMOEA models to optimize WLA.

  • WLAs are produced under different AOGCMs and emission scenarios.

  • Ranges of the uncertainties lead to a polluted dam reservoir with higher probability are discovered by PRIM and CART methods.

Abstract

In recent years, uncertainties in climate change have made the allocation of waste load a severe challenge. This research develops a new model framework integrating a soil and water assessment tool (SWAT) for hydrological simulation of the river basin, Borg many-objective evolutionary algorithm (Borg MOEA) for optimization, and a software for many-objective robust decision making (OpenMORDM) to minimize these uncertainties and identifying the most robust waste load allocation plans resisting the impact of system uncertainties. This framework has been applied to the Golgol River basin, Iran, which is the main source of the Ilam dam water supply. In this study, first, temperature and precipitation were predicted in 2080–2099. Then, The SWAT model was used to simulate runoff and biochemical oxygen demand (BOD). Afterward, by specifying the objective functions, the Borg MOEA produced the most optimal waste load allocation scenarios in the driest month of the future (August), So that the amount of BOD in the dam reservoir remains less than 5 mg/L. In the next step, the robustness of the solutions analyzed using OpenMORDM. Then, the results were analyzed using the patient rule induction method (PRIM) and classification and regression tree (CART) methods to identify the system vulnerability domains and minimizing false positives. Through the analysis, four climate scenarios have been used to explain how the methodological selection of them affects the resulting candidate planning solutions. The results indicated that individual design selections from the water managers or researchers could influence the efficacy of the model framework, which in turn results in failure of the most robust waste-load allocation strategies under future climate change.

Introduction

Lately, various approaches have been developed to address the uncertainties associated with water resources management. Looking for the most robust design alternatives that minimize deviation from future plans is one of the most critical elements in water resources management. This challenge is exacerbated by deep uncertainties so that the set of all possible future events, as well as their associated probability distributions themselves, contain uncertainties. Though, many robustness frameworks have expanded lately include: information gap decision theory (Ben-Haim, 2004), decision scaling (Brown et al., 2012, Brown, 2010), (many-objective) Robust Decision Making (Lempert, 2002, Kasprzyk et al., 2013, Lempert et al., 2013, Lempert et al., 2006, Lempert and Collins, 2007, Groves and Lempert, 2007, Weaver et al., 2013).

However, various researches have studied and organized ordinary methodological elections in robustness-focused decision support frameworks (Dittrich et al., 2016, Herman et al., 2015, Kwakkel and Haasnoot, 2019, Maier et al., 2016).

Among all these approaches, MORDM is vastly superior in implementing interactive visual analytics to distinguish proper answers, trade-offs recognition among various plans, and evaluate their operation under the influence of deep uncertainties. This method has so far been used to solve several water management issues (Singh et al. 2015).

Problem formulation is the first stage in MORDM, in which the operation goals, the specific uncertainties, and the computational model are received from the beneficiaries. Furthermore, critical constraints or other assumptions of the operation of the system are expressed. In the next step, the previous stage becomes a multi-objective search to create design alternatives for the problem.

For determining the robust performance rules of a water supply system, Kang Ren et al. (2019) utilized the MORDM framework. The uncertainty analysis and multi-objective optimization tools were employed first to show the trade-offs between the conflicting objectives and, second, reveal the sensitive factors for the possible states of the system. Results determined that the robust operating rules offer significant insights into the hurdles posed by deep uncertainties and provide a management template for decision making on climate change and complicated human activities.

Moallemi et al. (2019) examined two frameworks: Epoch–Era Analysis and Robust Decision Making. They deduced that the supremacy of each framework is not inevitable, and depends on the circumstances and the problem. While Robust Decision Making is a more impressing framework under contested and obscure information about the future, Epoch–Era Analysis can operate better when stakeholder knowledge about the future provides a deliberative specification of future scenarios.

Trindade et al. (2017) developed a multi-stakeholder MORDM framework to better account for deeply uncertain factors when recognizing collaborative drought management strategies. Their results revealed that appropriately designing adaptive risk-of-failure action triggers needed stressing them with a broad sample of deeply uncertain factors in the computational search condition of MORDM. The results from their research had a general benefit for areas where neighboring municipalities can profit from joint regional water portfolio planning.

Newly, Multi-Objective Evolutionary Algorithms (MOEAs) (Reed et al., 2013) have been considered in water resources management. These algorithms have been advanced to obtain the most suitable trade-off answers. Instead of gaining a solution that can optimize all objectives concurrently (Coello Coello et al., 2007), MOEAs can explore multiple Pareto-optimal results in a profoundly complicated search space because of their capability to employ the identities of recombinant answers (Zitzler and Thiele, 1999).

For a complex allocation system, MOEAs provide the optimization output as an approximate set of Pareto answers, instead of a single best solution. As managers' portfolio preferences diverged as trade-off information increased (Smith et al., 2019), choosing the most robust solutions among these alternatives creates new challenges for policymakers.

Previous studies have employed a variety of methods to select the robust alternatives in water resources management, among them are geometric angle-based pruning algorithm (Sudeng and Wattanapongsakorn, 2014), visually interactive decision-making and design using evolutionary multi-objective optimization (VIDEO) (Kollat and Reed, 2007) and MORDM (Kasprzyk et al., 2013).

Yan et al. (2017) advanced a model framework combining MORDM and biophysical modeling. Their case study was the Pearl River basin (PRB) China, where to reduce saltwater intrusion in the dry season, adequate flow to the delta was needed. Before recognizing and evaluating robust water allocation plans for the future, the performance of ten ultra-modern MOEAs was assessed for allocating water in the PRB The Borg MOEA, which is a self-adaptive optimization algorithm, had superior achievement through the historical periods. Therefore, it was elected to create new water allocation plans for the future (2079–2099). The study noted that the framework could work inadequately because of unpredictable climate change influences on water resources.

The purpose of this paper is to advance a model framework that combines many-objective robust decision making with hydrological modeling to identify robust waste load allocation plans under climate change. Through these analyses, four climate scenarios have been adopted to demonstrate how the methodological choice of them changes the resulting candidate planning solutions.

In this study, the OpenMORDM is utilized, which is an open-source application of MORDM in the R programming language. OpenMORDM provides a platform for constructive decision support, enabling analysts to interactively identify promising options and possible vulnerabilities while balancing contrasting objectives (Hadka et al., 2015). R language was chosen because of its widespread use in environmental modeling, and access to pre-built analytical and statistical packages selected.

In addition to the preceding, forecasting future climate change is also subject to uncertainty. (Dessai and Hulme, 2007). One of the effects of climate change is on hydrological cycles which, result in changing flow and discharge of rivers. These factors cause changes in the transmission and distribution of pollutants in the environment and generally affect the quantity and quality of water. All of these changes will introduce uncertainties that will pose significant challenges for water resource policymakers and managers.

For example, when 20 General Circulation Models (GCMs) were employed to create 39 runs of the 21st-century for the Murray-Darling Basin in Lim and Roderick (2009) study, the results indicated that 22 runs showed growth in yearly mean precipitation by the end of the 21st century, while 17 runs were showing this decline.

Robust optimization was practiced by Mortazavi-Naeini et al. (2015), to protect bulk water supply quality in an urban area against drought and climate change uncertainties and measured the amount of potential evapotranspiration (PET) and future rainfall. For the 23 GCMs obtained from a previous study CSIRO-BoM (2007), the Multi-Site Stochastic Model was used to generate 10,000 50-year replicates of daily precipitation and PET based on these values. It should be noted that just one emission scenario (A1F1) was involved in their study.

Culley et al. (2016) expanded a bottom-up framework to determine the maximum operational acceptance capacity of water resource systems regarding the space of future climate exposure which was based on seven GCMs and six regional climate models under three representative concentration pathways (RCPs).

There is no agreement on what will occur in the future climate change, and this will cause problems with making decisions for managing water resources efficiently.

Uncertainty about future weather forecasts is unlikely to diminish significantly in the near future. For managing water resources under uncertainty in climate change, it is essential to implement forecasting for a variety of emission scenarios resulting from multiple GCMs (Pierce et al., 2009, Teutschbein et al., 2015).

The remainder of the paper is structured as follows. In Section 2, the model framework has been defined, including models and datasets utilized in this paper. The Golgol River basin, Iran, has been presented in Section 3. As a case study. Conclusion in Section 4 contains lessons learned from this study and proposals for the future.

Section snippets

Model framework

Fig. 1 depicted the model framework combining various modeling tools and data employed in this research. This complex framework has a significant role as a tool to facilitate robust waste load allocation in the Golgol River basin, which includes a model for generating random climate data (Lars-WG), a hydrological model (SWAT), a state-of-the-art optimization algorithm (Borg MOEA), and an open-source software (OpenMORDM).

To study the effects of climate change on various systems in future

Case study

The study area in this research is Golgol watershed with an area of 279.53 km2 in the Ilam province, Iran. It is one of the subbasins of the Ilam Dam, and it is located between 46° 16′ 36′' and 46° 38′ 32′' eastern longitudes and between 33° 23′ 46″ and 33° 38′ 12′' northern latitudes (Fig. 2). The maximum height of the basin is 2065 m, its minimum height is 1052 m, and the average height is 1578 m above sea level. The average slope of the basin is 39.5%, and the average precipitation of the

Formulating the optimization problem

From Lars-WG and SWAT outputs, it has been observed that because of climate change, river flow (Fig. 3), and rainfall (Fig. 4) of the dry season are likely to decrease in the future within all climate scenarios. This may result in decreasing the water quality of the dam reservoir by the increase in the concentration of BOD discharge from upstream stakeholders (Table 1), so seeking an appropriate pollution strategy for BOD discharge to prevent further contamination of the dam reservoir has led

Conclusion

The model framework presented in this study combines MORDM with hydrological modeling to recognize robust waste load allocation strategies under four climate scenarios in the future, and it contains several parts.

First, the selected AOGCM data were down-scaled by the Lars-WG statistical model in order to predict temperature and precipitation based on climate change's effects. The utilized models in this study are HADCM3 and ECHAM5-OM, which are applied under two scenarios (A2 and B1) and in one

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (48)

  • M. Mortazavi-Naeini et al.

    Robust optimization to secure urban bulk water supply against extreme drought and uncertain climate change

    Environ. Model. Softw.

    (2015)
  • P.M. Reed et al.

    Evolutionary multi-objective optimization in water resources: the past, present, and future

    Adv. Water Resour.

    (2013)
  • K. Ren et al.

    Defining the robust operating rule for multi-purpose water reservoirs under deep uncertainties

    J. Hydrol.

    (2019)
  • R. Smith et al.

    Testing the potential of Multi-objective Evolutionary Algorithms (MOEAs) with Colorado water managers

    Environ. Model. Softw.

    (2019)
  • B. Trindade et al.

    Reducing regional drought vulnerabilities and multi-city robustness conflicts using many-objective optimization under deep uncertainty

    Adv. Water Resour.

    (2017)
  • D. Yan et al.

    Many-objective robust decision making for water allocation under climate change

    Sci. Total Environ.

    (2017)
  • L. Breiman et al.

    Classification and Regression Trees

    (1984)
  • C. Brown et al.

    Decision scaling: linking bottom-up vulnerability analysis with climate projections in the water sector

    Water Resour. Res.

    (2012)
  • Brown, C., 2010. Decision-scaling for Robust Planning and Policy under Climate Uncertainty. World Resources Report...
  • C. Coello Coello et al.

    Evolutionary Algorithms for Solving Multi-objective Problems

    (2007)
  • CSIRO-BoM, 2007. Climate Change in Australia: Technical Report....
  • S. Culley et al.

    A bottom-up approach to identifying the maximum operational adaptive capacity of water resource systems to a changing climate

    Water Resour. Res.

    (2016)
  • J.H. Friedman et al.

    Bump hunting in high-dimensional data

    Stat. Comput.

    (1999)
  • D. Hadka et al.

    Diagnostic assessment of the Borg MOEA on many-objective product family design problems. WCCI world congress on computational intelligence

    Congr. Evol. Comput.

    (2012)
  • Cited by (11)

    • Multi-objective robust decision-making for LIDs implementation under climatic change

      2023, Journal of Hydrology
      Citation Excerpt :

      Results indicated that using MORDM can improve the performance of optimal operating rules which were previously unable to deal with uncertainties. Meysami and Niksokhan (2020) used a framework integrating SWAT and MORDM to identify the most robust waste load allocation strategies resisting the system uncertainty. Employing this method under different climate scenarios revealed that the model's efficacy highly depends on the managers'/stakeholders' design preferences and methodological selection of the climate scenarios.

    • Increase thermal conductivity of aqueous mixture by additives graphene nanoparticles in water via an experimental/numerical study: Synthesise, characterization, conductivity measurement, and neural network modeling

      2020, International Communications in Heat and Mass Transfer
      Citation Excerpt :

      Y. Xu et al. [12] and I. Kazemi et al. [13] applied ANN on the calculated correlations and reported the fit improvement between correlation and empirical data. Further, in another study, R. Meysami et al. [14] developed Borg many-objective evolutionary algorithm (Borg MOEA) for optimization, which may lead the next generation of nanofluid modeling. In this study, by the “Top-down method”, Graphene has been produced.

    View all citing articles on Scopus
    View full text