Systems Modelling, Optimization and Decision Support

Introduction and Context

Water resources management is fundamentally a systems problem. Rivers, reservoirs, aquifers, irrigation schemes, cities, industries, ecosystems and institutions are interconnected through physical flows, economic relationships and governance rules. An intervention in one part of the system may produce benefits in one location and adverse effects elsewhere. For example, increasing reservoir releases can improve downstream water supply while reducing hydropower generation and storage security. Similarly, groundwater pumping may sustain irrigation in the short term but cause long-term depletion, land subsidence or saline intrusion.

Systems modelling provides a structured representation of these interactions. Optimization identifies preferred decisions under specified objectives and constraints. Decision-support systems combine models, data, stakeholder preferences and visualization to assist—not replace—professional judgement. At postgraduate level, the central issue is not simply how to solve a mathematical model, but how to formulate a defensible decision problem, represent uncertainty, assess trade-offs and communicate the consequences of alternative policies.

This chapter considers simulation, optimization, multi-criteria analysis, scenario planning, sensitivity analysis and robust decision-making for water allocation and operations.

Learning Objectives

By the end of this chapter, students should be able to:

1. Explain the role and structure of systems models in water resources management.

2. Formulate mass-balance simulation models for reservoirs and allocation systems.

3. Distinguish simulation, optimization and decision-support applications.

4. Formulate linear, nonlinear, dynamic and multi-objective optimization problems.

5. Apply multi-criteria decision analysis to competing water-management alternatives.

6. Construct scenarios and evaluate sensitivity, uncertainty and robustness.

7. Interpret model outputs critically and communicate them to stakeholders.

8. Recognize ethical, institutional, computational and governance limitations.

1. Water Management as a System

A system consists of components, relationships, inputs, outputs, feedbacks and boundaries. In water resources, typical components include:

• State variables: reservoir storage, groundwater head, soil moisture and accumulated sediment.
• Inputs: precipitation, inflow, recharge, return flow and external water transfers.
• Control variables: releases, pumping, diversion, irrigation scheduling and operating rules.
• Outputs: water deliveries, hydropower, flood damage, ecological flows and financial costs.
• Constraints: storage capacity, channel capacity, legal entitlements, environmental standards and energy limits.
• Performance indicators: reliability, resilience, vulnerability, equity, economic value and ecological condition.

A model boundary determines what is represented explicitly and what is treated as external. A basin-scale model may represent reservoirs and irrigation districts but treat climate as an input. A broader integrated model may include energy prices, land use and climate feedbacks. Boundaries should be chosen according to the decision being studied, not merely according to data availability.

Water systems contain feedback. Low reservoir storage may trigger conservation restrictions, which reduce demand and alter future storage. High groundwater pumping may lower the water table, increasing pumping costs and reducing future abstraction. These feedbacks make static averages inadequate for many strategic decisions.

Important Modelling Terminology

A conceptual model describes system structure and causal relationships. A mathematical model expresses these relationships using equations, inequalities or logical rules. A simulation model evaluates system behaviour for specified inputs and decisions. An optimization model searches for decisions that maximize or minimize an objective subject to constraints. A calibration process estimates parameters using observed data, whereas validation evaluates performance using independent data or conditions. Verification asks whether the model has been implemented correctly; validation asks whether it is an adequate representation for the intended use.

2. Simulation Models for Allocation and Operations

Simulation answers the question: “What happens if this policy is followed?” It does not automatically identify the best policy. Simulation is particularly suitable when operating rules are complex, nonlinear or institutionally prescribed.

2.1 Reservoir mass balance

For a reservoir evaluated over time step t, the continuity equation is

\[

S_{t+1}=S_t+I_t+P_t-R_t-E_t-L_t,

\]

where:

• \(S_t\) and \(S_{t+1}\) are beginning and end storage volumes;
• \(I_t\) is inflow volume;
• \(P_t\) is direct precipitation on the reservoir;
• \(R_t\) is controlled release;
• \(E_t\) is evaporation loss;
• \(L_t\) represents other losses, such as seepage;
• all terms must use the same volume units and time interval.

The equation assumes that storage is adequately represented by a single state variable and that all significant inflows and outflows are included. It is a conservation equation, not an operating policy. A simulation additionally requires rules such as:

\[

R_t=\min\{R^{\text{demand}}_t,\; S_t+I_t+P_t-E_t-L_t-S^{\min}\},

\]

subject to nonnegative release and a required minimum end-of-period storage \(S^{\min}\). This illustrative rule prioritizes demand while protecting storage. Real rules may include flood-control zones, hydropower requirements, environmental releases and priority classes.

Storage must also satisfy

\[

S^{\min}\leq S_{t+1}\leq S^{\max}.

\]

If the calculated storage exceeds \(S^{\max}\), the excess is spill or an additional outflow. If it falls below \(S^{\min}\), the policy is infeasible unless shortage or emergency drawdown is permitted.

2.2 Allocation and return flows

For a basin with users i, a simple allocation balance is

\[

A_t=\sum_i D_{i,t},

\]

where \(D_{i,t}\) is delivery to user i. If the available supply is less than total demand, a rationing rule determines actual delivery \(W_{i,t}\). A proportional shortage rule is

\[

W_{i,t}=\alpha_tD_{i,t}, \qquad \alpha_t=\min\left(1,\frac{A_t}{\sum_iD_{i,t}}\right).

\]

This rule is transparent but may conflict with legal priorities or equity requirements. Return flow can be represented as \(q_iW_{i,t}\), where \(q_i\) is the fraction returning to the water system. The assumption that return-flow fractions are constant should be tested because they may vary with soil, crop, season and groundwater conditions.

2.3 Performance measures

Common simulation indicators include:

• Reliability: proportion of periods in which a specified target is met.
• Resilience: probability that the system returns to satisfactory performance after failure.
• Vulnerability: severity of failure, often measured by deficit magnitude or duration.
• Shortage volume: cumulative unmet demand.
• Environmental compliance: proportion of periods meeting ecological-flow requirements.

These measures answer different questions. A system can have high reliability but severe occasional failures, or frequent small failures but low vulnerability. Consequently, no single indicator is sufficient for strategic evaluation.

3. Optimization Formulations

Optimization identifies a feasible decision vector \(x\) that improves an objective function. The general form is

\[

\text{maximize or minimize } f(x)\quad\text{subject to}\quad g_j(x)\leq 0,\; j=1,\ldots,m,

\]

with bounds or other restrictions on \(x\). The objective and constraints must reflect the decision horizon, legal obligations and system physics.

3.1 Linear programming

A linear program has a linear objective and linear constraints:

\[

\max Z=\sum_{i=1}^{n}c_ix_i

\]

subject to

\[

\sum_{i=1}^{n}a_{ji}x_i\leq b_j,\quad j=1,\ldots,m,\qquad x_i\geq0.

\]

Here \(x_i\) are decision variables, \(c_i\) are objective coefficients, \(a_{ji}\) are resource-use coefficients and \(b_j\) are resource limits. Linear programming is useful for seasonal allocation, crop planning, reservoir releases and water-quality load allocation when relationships can reasonably be approximated as linear.

For example, if crops A and B require 3 and 5 units of water per hectare, respectively, and only 120 units are available, the water constraint is \(3x_A+5x_B\leq120\). Other constraints may represent land, labour, minimum production or policy requirements. Linear programming is computationally efficient and provides shadow prices, which estimate the marginal value of relaxing a binding constraint under model assumptions.

3.2 Nonlinear and dynamic optimization

Water systems often contain nonlinear relationships. Hydropower benefit, for example, depends approximately on

\[

E_t=\eta\rho g R_t H_t,

\]

where \(E_t\) is energy, \(\eta\) is efficiency, \(\rho\) is water density, \(g\) is gravitational acceleration, \(R_t\) is turbine flow and \(H_t\) is hydraulic head. Because head varies with storage, the objective may be nonlinear. The equation assumes consistent SI units and neglects detailed turbine operating limits.

A dynamic optimization model links decisions across time. A simplified deterministic reservoir formulation is

\[

\max_{R_t,S_t}\sum_{t=1}^{T}\beta^{t-1}B_t(R_t,S_t)-C_t(R_t,S_t),

\]

subject to the mass balance and storage constraints for every period. Here \(B_t\) is benefit, \(C_t\) is cost and \(\beta\) is a discount factor. Discounting is appropriate for economic comparison over time, but it can undervalue long-term ecological benefits; the discount rate should therefore be justified rather than selected mechanically.

3.3 Stochastic and robust optimization

A deterministic model uses one assumed sequence of inflows or demands. A stochastic model represents alternative states of nature, often with probabilities. A two-stage stochastic formulation permits decisions before uncertainty is known and recourse decisions after information becomes available. Its expected-value objective may be written as

\[

\max \sum_{s\in\Omega}p_s F(x,y_s),

\]

where \(x\) are here-and-now decisions, \(y_s\) are scenario-dependent decisions, \(p_s\) are scenario probabilities and \(F\) is scenario performance.

Probabilities are often highly uncertain in climate-sensitive basins. Robust optimization therefore seeks decisions that perform acceptably across a set of plausible conditions rather than maximizing expected performance under one probability distribution. A typical robust objective is

\[

\max_x \min_{s\in\Omega}F(x,s),

\]

which maximizes worst-case performance. Strict worst-case optimization can be excessively conservative; practical approaches may instead limit maximum regret, impose reliability requirements across scenarios or use satisficing thresholds.

4. Multi-Objective Decision-Making

Water decisions rarely have one objective. Common objectives include urban reliability, agricultural production, hydropower, ecological protection, flood reduction, affordability and intergenerational equity. These objectives may be conflicting and measured in incompatible units.

A weighted-sum formulation is

\[

\max U(x)=\sum_{k=1}^{K}w_k\hat f_k(x),

\]

where \(f_k(x)\) is criterion k, \(\hat f_k\) is a normalized value and \(w_k\) is a nonnegative weight with \(\sum w_k=1\). Normalization is essential because cubic metres, dollars and ecological scores cannot be added directly. A min–max normalization for a benefit criterion is

\[

\hat f_k(x)=\frac{f_k(x)-f_k^{\min}}{f_k^{\max}-f_k^{\min}},

\]

where the reference minimum and maximum must be defined consistently. For a cost criterion, the direction is reversed.

The weighted sum is transparent but has limitations. Weights may conceal value judgements, results can depend strongly on normalization and some criteria may be non-compensatory. An alternative is the epsilon-constraint method, in which one objective is optimized while other objectives are expressed as minimum performance constraints. The set of non-dominated solutions forms the Pareto frontier: a solution is Pareto-dominated if another feasible solution is at least as good on every criterion and better on at least one.

Multi-criteria decision analysis (MCDA) generally includes:

1. defining alternatives and criteria;

2. specifying criterion direction and measurement;

3. screening infeasible alternatives;

4. normalizing or otherwise transforming scores;

5. eliciting weights or thresholds;

6. aggregating scores or applying outranking;

7. testing sensitivity to weights and scores;

8. documenting stakeholder disagreement.

Methods such as AHP, TOPSIS and ELECTRE can support structured comparison, but no method eliminates the underlying normative choices. Results should be presented as conditional on assumptions, not as objectively inevitable rankings.

5. Scenario Planning and Uncertainty

Uncertainty arises from natural variability, incomplete observations, parameter error, structural error and future changes in climate, population, technology, policy and economics. It is useful to distinguish:

• Aleatory uncertainty: inherent variability, such as random rainfall.
• Epistemic uncertainty: imperfect knowledge, such as uncertain aquifer parameters.
• Deep uncertainty: disagreement or lack of confidence about models, probabilities or future conditions.

Scenario planning explores coherent combinations of assumptions. A scenario is not necessarily a forecast. It may represent, for example, rapid urban growth, prolonged drought, altered monsoon timing, energy-price change or a new environmental regulation.

A sound scenario process includes a baseline, stress scenarios and transformative scenarios. Each scenario should specify assumptions, time horizon, spatial scale, data sources and indicators. Double counting must be avoided: if a climate scenario already includes a change in precipitation, that change should not be added again through an independent adjustment unless the modelling design explicitly requires it.

Sensitivity analysis

Sensitivity analysis tests how outputs change when inputs or assumptions vary. One-at-a-time analysis changes one parameter while holding others constant and is easy to explain, but it may miss interactions. Global methods vary several uncertain inputs simultaneously and may use sampling, variance decomposition or scenario ensembles.

For a model output \(Y=f(x)\), a local sensitivity to parameter \(x_i\) can be approximated by

\[

S_i\approx\frac{\Delta Y/Y}{\Delta x_i/x_i},

\]

provided the baseline values are nonzero and perturbations are sufficiently small. This is a dimensionless elasticity-like measure, not a universal causal coefficient. A large sensitivity indicates that improved knowledge or careful governance of that parameter may be valuable.

6. Decision-Support Systems and the Modelling Workflow

A decision-support system typically integrates a database, analytical models, scenario tools, visualization and a user interface. Geographic information systems can display spatial demand, groundwater decline, flood exposure and ecological assets. Dashboards can show storage trajectories, forecast inflows and shortage indicators, but visual simplicity must not conceal uncertainty.

A defensible workflow is:

1. Frame the decision: identify actors, choices, objectives, constraints, time horizon and consequences.

2. Develop a conceptual model: map stocks, flows, feedbacks and causal assumptions.

3. Collect and audit data: record units, temporal resolution, missing values, provenance and quality.

4. Select model structure: use the least complex model that can answer the decision question adequately.

5. Formulate and implement: encode equations, rules, constraints and objective functions.

6. Verify: test mass balance, boundary conditions, extreme cases, units and code logic.

7. Calibrate and validate: compare with observations, using appropriate performance measures and independent periods where possible.

8. Generate alternatives and scenarios: include current practice, feasible improvements and stress conditions.

9. Evaluate trade-offs and robustness: examine performance, uncertainty, sensitivity and distributional effects.

10. Engage stakeholders and document assumptions: revise the model where legitimate knowledge or values have been omitted.

11. Monitor and adapt: compare observed outcomes with predictions and update the model or policy.

Model validation is purpose-specific. A model adequate for monthly reservoir yield may be unsuitable for flood routing or ecological habitat analysis. Calibration should not be used to conceal structural inadequacy through excessive parameter adjustment.

7. Worked Example: Seasonal Reservoir Allocation

Consider an illustrative one-period reservoir at the beginning of a dry season. Initial storage is 80 million m³, inflow is 30 million m³, evaporation and seepage together are 5 million m³, and the required end-of-season reserve is 20 million m³. Urban demand is 35 million m³ and irrigation demand is 45 million m³. The reservoir capacity is 100 million m³.

Available water for release while preserving the reserve is

\[

80+30-5-20=85\text{ million m}^3.

\]

Total demand is 80 million m³, so both demands can theoretically be fully met and end storage is

\[

S_1=80+30-5-35-45=25\text{ million m}^3.

\]

The reserve is satisfied with a surplus of 5 million m³. This is a feasible simulation result under the stated assumptions.

Now suppose inflow falls to 10 million m³. Available release while preserving the reserve becomes

\[

80+10-5-20=65\text{ million m}^3.

\]

Using proportional rationing,

\[

\alpha=\frac{65}{80}=0.8125.

\]

Urban delivery is \(35(0.8125)=28.4375\) million m³ and irrigation delivery is \(45(0.8125)=36.5625\) million m³. End storage is 20 million m³. The result illustrates that proportional rationing is feasible but not necessarily equitable or legally acceptable. If urban supply has priority, an alternative rule might meet the full urban demand and allocate the remaining 30 million m³ to irrigation. The two rules have identical total release but different distributional consequences.

An optimization model could select deliveries \(W_U\) and \(W_I\) by maximizing weighted benefits:

\[

\max \; b_UW_U+b_IW_I

\]

subject to

\[

W_U+W_I\leq65,\quad 0\leq W_U\leq35,\quad 0\leq W_I\leq45.

\]

The coefficients \(b_U\) and \(b_I\) must be justified; they may represent economic value, social priority or a policy-derived utility, and should not be interpreted as purely technical facts.

8. Applied Case: Illustrative Integrated Basin Planning Study

Background

The fictional Kalinga Basin contains a multipurpose reservoir, two irrigation districts, a rapidly growing city and a downstream wetland. Climate projections indicate greater variability in seasonal inflows, but the magnitude and timing are uncertain. Existing operations maximize annual irrigation deliveries and have no explicit drought trigger.

Problem

Recent droughts caused urban restrictions, reduced wetland flows and conflict between irrigation districts. The basin authority must choose a 20-year strategy involving reservoir rule changes, demand management, wastewater reuse and possible additional storage.

Analysis

The authority develops a monthly simulation model with reservoir continuity, urban demand, irrigation demand, evaporation, environmental-flow requirements and operating zones. Thirty climate and demand scenarios are generated, not as precise forecasts but as plausible stress combinations. Three candidate strategies are evaluated:

• S1: continue existing operations;
• S2: introduce drought triggers, leakage reduction and moderate reuse;
• S3: construct additional storage combined with reuse and environmental operating rules.

Performance indicators include urban reliability, irrigation shortage, wetland-flow compliance, capital cost, energy use and maximum annual shortage. An MCDA process is conducted separately with municipal, agricultural, ecological and community representatives. Rather than publish one ranking only, the authority presents Pareto-efficient strategies and a sensitivity analysis of weights.

Approach and solution

S1 performs well in average historical conditions but fails under prolonged drought. S3 improves reliability in several scenarios but has high financial and environmental costs and a long construction lead time. S2 produces slightly lower average reliability than S3 but substantially improves performance in dry scenarios, can be implemented incrementally and has lower irreversible commitment.

A robust decision rule selects S2 as the initial strategy, with staged investments and monitoring thresholds. If storage reliability falls below a specified trigger for several consecutive years, additional reuse capacity is authorized. The plan includes annual review of environmental-flow compliance and five-year reassessment of demand and climate information.

Outcome and learning points

The case demonstrates that the preferred strategy is not necessarily the one with the highest mean benefit. Flexibility, reversibility, distributional effects and performance under stress may be more important than average results. It also shows why stakeholder participation is substantive: the selection of criteria and thresholds changes the decision, not merely its presentation.

9. Limitations, Ethics and Governance

Models can create false confidence when precise numerical outputs conceal uncertain assumptions. Key limitations include poor data, nonstationary climate, omitted feedbacks, inappropriate spatial or temporal resolution, parameter equifinality and computational simplification. Optimization may exploit loopholes in a formulation—for example, meeting an environmental-flow constraint at a politically unacceptable time or location.

Ethical modelling requires transparency about who benefits and who bears risk. Aggregate economic optimization can disadvantage low-income users, customary water holders or ecosystems that lack market prices. Distributional indicators, minimum service standards and rights-based constraints may be needed. Stakeholders should have access to assumptions and meaningful opportunities to challenge them.

Governance requirements include compliance with water-allocation law, environmental regulation, public procurement rules, data protection and professional standards. A model should not be used to evade statutory consultation or to convert a political choice into an apparently neutral technical result. Version control, audit trails, reproducible data and documented approval processes are essential for high-consequence decisions.

10. Technology, AI and Contemporary Developments

Cloud computing, remote sensing, digital twins, ensemble forecasting and open-source optimization libraries are expanding modelling capability. Machine learning can improve demand forecasting, detect anomalies and emulate computationally expensive simulations. However, predictive accuracy does not guarantee causal validity or policy suitability. Black-box predictions may be difficult to audit, may reproduce historical inequities and may fail under climate conditions outside the training data.

A prudent approach is hybrid modelling: combine physically based mass-balance and hydraulic models with data-driven components where they add demonstrable value. AI-generated recommendations should be independently verified, uncertainty should be quantified where possible and human decision-makers should retain accountability. Digital twins should be treated as continuously updated decision-support representations, not perfect replicas of reality.

Best-Practice Principles

1. Begin with the decision and stakeholders, not with a preferred algorithm.

2. Use conservation laws and transparent operating rules wherever applicable.

3. Keep units, time steps and spatial boundaries explicit.

4. Separate calibration, validation and verification.

5. Test extreme conditions and inspect mass-balance closure.

6. Represent uncertainty through scenarios or ensembles rather than one forecast.

7. Report Pareto trade-offs and sensitivity to weights.

8. Include equity, ecological integrity and legal obligations explicitly.

9. Prefer adaptive and reversible strategies under deep uncertainty.

10. Document data provenance, code, assumptions and changes.

11. Communicate ranges, failure modes and confidence limits—not only point estimates.

12. Monitor outcomes and update both model and policy.

Chapter Summary

Systems modelling represents water resources as interconnected stocks, flows, decisions, constraints and feedbacks. Simulation evaluates the consequences of specified operating rules, while optimization searches for preferred decisions subject to physical, economic, legal and environmental constraints. Reservoir mass balance is the foundation of many allocation models, but realistic applications also require demand rules, priorities, return flows and performance indicators.

Linear programming is efficient for allocation problems with linear relationships; nonlinear and dynamic models address head-dependent hydropower, intertemporal storage and other complex relationships. Stochastic and robust approaches recognize uncertain inflows, demand and climate. Multi-criteria analysis makes competing objectives explicit, while Pareto analysis reveals trade-offs that a single weighted score may conceal. Scenario planning and sensitivity analysis test whether conclusions remain credible under alternative futures and assumptions.

Decision-support systems integrate data, models, stakeholder preferences and visualization. Their quality depends not only on algorithms but also on model framing, data governance, verification, validation, transparency and institutional legitimacy. Ethical practice requires attention to distributional impacts, ecological rights and accountability. Under deep uncertainty, adaptive strategies that perform acceptably across plausible futures are often preferable to rigid solutions optimized for one forecast.

Key Terms

Systems model: A structured representation of components, relationships, inputs, outputs and feedbacks in a system.

Simulation: Evaluation of system behaviour under specified inputs, decisions and operating rules.

Optimization: Systematic search for the best feasible decision according to a defined objective.

State variable: A quantity, such as storage, that describes the condition of a system at a given time.

Control variable: A decision variable used to operate or manage a system.

Mass balance: Conservation relationship equating initial stock plus inflows minus outflows to final stock.

Linear programming: Optimization with linear objective and constraint relationships.

Dynamic optimization: Optimization in which decisions and states are linked across time.

Stochastic optimization: Optimization that represents uncertain outcomes or scenarios with probabilities.

Robust optimization: Selection of decisions that perform acceptably across uncertain conditions.

Multi-criteria decision analysis: Structured comparison of alternatives using several criteria.

Pareto frontier: Set of feasible solutions that are not dominated by another solution.

Sensitivity analysis: Investigation of how results change when assumptions or inputs vary.

Scenario: A coherent, plausible set of assumptions about future conditions.

Reliability: Frequency or probability with which a required performance target is met.

Resilience: Ability of a system to recover after a failure.

Vulnerability: Severity of consequences when a system fails.

Verification: Checking whether a model has been implemented correctly.

Validation: Assessing whether a model is adequate for its intended purpose.

Decision-support system: Integrated data, modelling and communication tools that assist decisions.

Review Questions

Short Answer Questions

1. Distinguish simulation from optimization in water resources management.

2. State the reservoir mass-balance equation and define its terms.

3. What is meant by a Pareto-efficient solution?

4. Differentiate aleatory, epistemic and deep uncertainty.

5. Why are verification and validation separate activities?

Long Answer Questions

1. Explain how a dynamic reservoir optimization model can represent competing urban, irrigation, hydropower and ecological objectives.

2. Discuss the advantages and limitations of weighted-sum MCDA and the epsilon-constraint method.

3. Describe a complete workflow for developing and implementing a basin-scale decision-support system.

4. Explain how robust decision-making differs from expected-value stochastic optimization.

5. Critically assess the ethical and governance risks of using optimization models for water allocation.

Analytical / Application Questions

1. A reservoir begins with 60 million m³, receives 25 million m³ inflow, loses 5 million m³ to evaporation, and must retain 15 million m³. If urban and irrigation demands are 30 and 40 million m³, respectively, determine the available release and propose two defensible rationing policies.

2. Design an MCDA framework for selecting among leakage reduction, desalination and a new reservoir. Specify at least six criteria, their directions, a normalization approach and a sensitivity analysis.

3. Develop a scenario-based evaluation plan for a basin facing uncertain climate change, population growth and environmental regulation. Explain how you would identify a robust and adaptive strategy.

Multiple Choice Questions with Answers

Question 1

Which equation represents a reservoir continuity relationship?

A. S_{t+1}=S_t+inflows−outflows

B. S_{t+1}=S_t×inflows×outflows

C. S_{t+1}=inflows−storage capacity

D. S_{t+1}=outflows−S_t

Correct Answer: A. S_{t+1}=S_t+inflows−outflows

Question 2

What is the principal purpose of simulation?

A. To guarantee an optimal policy

B. To evaluate system behaviour under specified decisions

C. To eliminate uncertainty from inputs

D. To replace stakeholder judgement

Correct Answer: B. To evaluate system behaviour under specified decisions

Question 3

In linear programming, what must be linear?

A. Only the data sources

B. The objective and constraints

C. Only the decision-maker preferences

D. The weather record

Correct Answer: B. The objective and constraints

Question 4

A Pareto-dominated solution is one for which:

A. No feasible solution exists

B. Another solution is at least as good on all criteria and better on one

C. All criteria have equal weights

D. The objective function is nonlinear

Correct Answer: B. Another solution is at least as good on all criteria and better on one

Question 5

Which approach is most directly associated with maximizing worst-case performance?

A. Robust optimization

B. Historical averaging

C. Simple proportional allocation

D. Calibration

Correct Answer: A. Robust optimization

Question 6

If available supply is 65 million m³ and total demand is 80 million m³, proportional rationing uses an allocation factor of:

A. 0.65

B. 0.8125

C. 1.2308

D. 15

Correct Answer: B. 0.8125

Question 7

Which activity checks whether computer code correctly implements the intended model?

A. Validation

B. Verification

C. Scenario planning

D. Weight elicitation

Correct Answer: B. Verification

Question 8

Why is normalization needed in a weighted-sum MCDA?

A. To make incompatible criterion units comparable

B. To remove all stakeholder values

C. To guarantee the highest-cost option wins

D. To convert every criterion into a legal constraint

Correct Answer: A. To make incompatible criterion units comparable

Question 9

Which is an example of a control variable in reservoir management?

A. Reservoir storage

B. Observed rainfall

C. Reservoir release

D. Historical evaporation

Correct Answer: C. Reservoir release

Question 10

Why may an adaptive strategy be preferred under deep uncertainty?

A. It ignores monitoring information

B. It remains adjustable as conditions and knowledge change

C. It always has the lowest initial cost

D. It requires no assumptions

Correct Answer: B. It remains adjustable as conditions and knowledge change