Digital and Emerging Environmental Engineering

Introduction and Context

Environmental engineering is entering a phase in which physical infrastructure is increasingly designed, operated and governed through digital systems. Water-treatment plants, sewer networks, solid-waste facilities, drainage systems and industrial environmental-control systems now generate large volumes of data from sensors, laboratory instruments, geographic information systems (GIS), building information modelling (BIM), supervisory control and data acquisition (SCADA) systems and enterprise platforms. At the same time, advanced membranes, resource-recovery processes, automation and artificial intelligence (AI) are changing what environmental infrastructure can achieve.

Digitalisation, however, is not simply the installation of sensors or software. It changes decision-making, professional responsibility, institutional power and the meaning of evidence. A prediction that is technically accurate may still be unsuitable if its data are biased, its uncertainty is hidden, its cybersecurity is weak or its recommendation cannot be explained to regulators and communities. The contemporary environmental engineer must therefore combine process knowledge with data engineering, systems thinking, risk management, ethics and life-cycle assessment.

This chapter examines digital and emerging technologies as socio-technical systems. The emphasis is not on technology adoption for its own sake, but on whether a technology improves environmental performance, resilience, equity, safety and accountability.

Learning Objectives

After studying this chapter, the learner should be able to:

1. Distinguish digital models, digital shadows and digital twins.

2. Explain BIM–GIS integration and its value for environmental assets.

3. Design an IoT monitoring architecture and evaluate data quality.

4. Describe automation, optimisation and machine-learning workflows for environmental systems.

5. Assess AI-assisted decisions using performance, uncertainty, ethics and governance criteria.

6. Explain the roles of advanced membranes and resource recovery in circular and net-zero systems.

7. Develop practical controls for data governance and cybersecurity.

8. Critically evaluate an integrated digital environmental-engineering application.

1. A Framework for Digital Environmental Engineering

1.1 From isolated instruments to cyber-physical systems

A traditional environmental facility is often represented by process-flow diagrams, design drawings, operating manuals and periodic laboratory results. A cyber-physical environmental system links physical processes to sensing, communications, computation and control. Its basic cycle is:

observe → transmit → interpret → decide → act → verify.

For example, online turbidity, ultraviolet transmittance and flow sensors may transmit data to a control platform. A model estimates treatment performance, an optimisation algorithm selects an operating condition, and actuators adjust coagulant dosing or membrane flux. Verification is essential: the result must be checked against independent measurements and safety constraints.

Digital systems should be specified by the decision they support. “Install more sensors” is not a sound objective. A better objective is: “Detect deterioration in membrane integrity early enough to prevent non-compliant water from entering storage.” This statement identifies the failure mode, time requirement, measurement need and regulatory consequence.

1.2 Levels of digital representation

A digital model is a representation used to analyse a system; it may operate independently of real-time data. A hydraulic model of a sewer network is an example.

A digital shadow is a digital representation that receives data from the physical system, generally in one direction. The physical system may not be automatically controlled by the representation.

A digital twin is a continuously connected digital representation whose state is updated by the physical asset and which can support prediction, simulation, diagnosis or controlled intervention. The term is sometimes used loosely. A 3D model alone is not a digital twin. A credible twin normally requires:

• a defined physical boundary and purpose;
• asset identity and consistent data structures;
• time-dependent operational data;
• validated process and failure models;
• mechanisms for uncertainty and data-quality management;
• controlled interfaces to decision-makers or actuators; and
• cybersecurity, version control and lifecycle governance.

A digital twin may be used for descriptive, diagnostic, predictive or prescriptive purposes. Descriptive twins show current conditions; diagnostic twins identify causes; predictive twins forecast future states; prescriptive twins compare actions subject to constraints.

2. BIM–GIS Integration for Environmental Infrastructure

2.1 Complementary representations

BIM represents detailed objects, geometry, materials, specifications, quantities and relationships within a facility or project. GIS represents assets and phenomena in geographic space, usually across wider areas and in relation to terrain, land use, catchments, property, hazards and populations. BIM is commonly detailed and facility-centred; GIS is geographically extensive and context-centred.

Integration supports questions such as:

• Which treatment assets are exposed to a flood hazard?
• Which sewer manholes are connected to a vulnerable receiving water?
• Which pipes installed with a particular material and age are failing in corrosive soil?
• How will a proposed pumping station affect energy use, noise and nearby communities?

Integration is difficult because coordinate systems, object classifications, levels of detail, identifiers, time references and data ownership differ. A practical architecture assigns each asset a persistent identifier and maintains a common data dictionary. Geometry should be transformed using documented coordinate-reference systems rather than visual approximation. Attributes such as installation date, material, capacity, inspection history and responsible organisation should be governed as carefully as geometry.

2.2 Lifecycle use

During planning, BIM–GIS supports site selection, environmental-impact analysis and options appraisal. During design and construction, it reduces clashes and improves quantity and change control. During operation, it connects asset registers, work orders, sensor locations and maintenance history. During decommissioning, it supports hazardous-material inventories and recovery planning.

The value is greatest when information requirements are defined before procurement. An organisation should specify who creates each data item, its acceptable accuracy, update frequency, retention period, access rights and verification method. Otherwise, an attractive model may become an expensive but unreliable visual archive.

3. IoT Monitoring and Data Quality

3.1 IoT architecture

An Internet of Things (IoT) monitoring system typically contains:

1. Sensing layer: measures variables such as flow, pressure, pH, dissolved oxygen, conductivity, temperature, methane or particulate matter.

2. Edge layer: performs filtering, time synchronisation, local storage, anomaly checks and sometimes immediate control.

3. Communication layer: uses wired networks, cellular links, radio, low-power wide-area networks or industrial protocols.

4. Platform layer: stores, contextualises and processes data.

5. Application layer: provides dashboards, alarms, forecasts, maintenance tools and regulatory reports.

6. Governance layer: controls identity, access, quality, retention, auditability and responsible use.

Sensor selection should begin with the variable required for a decision. Important specifications include range, resolution, accuracy, response time, drift, environmental tolerance, maintenance requirement, calibration method and failure behaviour. A low-cost sensor can be appropriate for screening but inappropriate for legal compliance unless validated against an approved method.

3.2 Data quality and uncertainty

Data quality is multidimensional. Common dimensions are accuracy, completeness, timeliness, consistency, validity, uniqueness and traceability. A complete data stream can still be wrong if a sensor is fouled or incorrectly installed.

Quality assurance should include calibration, field checks, plausibility limits, redundant measurements for critical variables, time-stamped maintenance records and independent laboratory comparison. Missing values must not automatically be replaced by zeros. Imputation, if used, should preserve a flag indicating that the value was estimated and should be excluded from compliance claims unless permitted by the applicable standard.

For a measured value x and a reference value x_ref, a simple relative error may be expressed as

relative error = |x − x_ref| / |x_ref|.

This expression is meaningful when x_ref is not zero and when the comparison is made under sufficiently similar conditions. It is a diagnostic measure, not a universal acceptance criterion. Acceptance limits must be established from instrument specifications, method validation and regulatory requirements.

3.3 Edge computing and resilient monitoring

Edge processing reduces communication demand and can maintain operation during network outages. It is appropriate for immediate alarms, interlocks and preliminary data screening. Cloud systems provide scalable storage and advanced analytics but introduce dependency on connectivity, service providers and cross-border data arrangements. Critical treatment safety should not depend solely on a remote cloud connection. A resilient design permits safe local operation, records events during outages and reconciles data after reconnection.

4. Automation, Optimisation and Control

4.1 Levels of automation

Automation ranges from alarms and operator decision support to closed-loop control. Environmental engineers should distinguish:

• monitoring: observing process variables;
• supervisory control: changing set points or schedules;
• regulatory control: maintaining a variable near a target, often through feedback;
• optimising control: selecting the best feasible operating policy;
• autonomous operation: allowing software to act within defined authority and safeguards.

Closed-loop autonomy is unsuitable where the process model is uncertain, consequences are severe or human intervention is legally required. Safety interlocks and independent protection layers should remain outside the optimisation software where practicable.

4.2 Optimisation formulation

An environmental operation can be formulated as a constrained optimisation problem:

Minimise J(u, x) subject to g(x, u) ≤ 0 and h(x, u) = 0,

where J is an objective function, u represents controllable decisions, x represents process states, g represents inequality constraints and h represents equality constraints. The objective may combine energy cost, chemical use, emissions, sludge production and risk. Constraints may include effluent limits, hydraulic capacity, equipment bounds and minimum disinfection performance.

The formulation is useful only when objectives and constraints are explicit. A solution that reduces energy while violating pathogen-control requirements is not an optimum in engineering terms. Multi-objective problems can be handled through weighted objectives, Pareto analysis or hierarchical priorities. Weights should be justified because they encode institutional and social values rather than merely mathematical preferences.

4.3 Model predictive control

Model predictive control (MPC) repeatedly uses a process model to forecast future behaviour over a finite horizon. At each interval it solves an optimisation problem, implements only the first action, obtains new measurements and repeats the calculation. MPC can manage interactions and constraints in aeration, pumping, reservoirs and advanced treatment. Its limitations include model error, sensor failure, computational requirements and the need for reliable fallback control.

5. Machine Learning and AI-Assisted Decision Support

5.1 Suitable tasks

Machine learning (ML) learns patterns from data for tasks such as classification, regression, anomaly detection, forecasting and image interpretation. Applications include predicting inflow, identifying leaks, forecasting algal blooms, classifying recyclable materials, estimating energy demand and detecting unusual treatment behaviour.

A responsible workflow includes:

1. define the decision and consequence of error;

2. assemble representative and legally usable data;

3. clean, label and document the data;

4. divide data into training, validation and independent test sets;

5. select a model appropriate to the task and available evidence;

6. evaluate performance against a meaningful baseline;

7. quantify uncertainty and examine failure modes;

8. validate under changing seasons, sites and operating regimes;

9. deploy with monitoring, human oversight and rollback procedures.

Randomly splitting time-series data can cause leakage: future information may inadvertently enter training data. Time-based or site-based validation is usually more realistic.

5.2 Performance and uncertainty

For a binary classifier, accuracy is the proportion of all correct predictions, but it can be misleading for rare events. Precision is the proportion of predicted positives that are truly positive; recall is the proportion of actual positives detected. A system for detecting hazardous contamination may prioritise recall, while an alarm system with costly false call-outs may require a balance between precision and recall.

For regression, mean absolute error is commonly defined as

MAE = (1/n) Σ |y_i − ŷ_i|,

where y_i is the observed value, ŷ_i is the predicted value and n is the number of observations. MAE has the same units as the predicted variable and is less dominated by large errors than squared-error measures. It does not by itself establish safety or generalisability.

Uncertainty should be communicated through prediction intervals, scenario ranges, confidence measures where statistically justified, or explicit abstention when the input is outside the training domain. An AI system should be able to say “insufficient evidence” rather than force a confident recommendation.

5.3 Explainability and human responsibility

Explainability may involve transparent models, feature importance, local explanations, counterfactuals or process-based hybrid models. Explanations must not be confused with proof of causation. Human operators remain responsible for decisions unless authority has been explicitly and lawfully delegated. A dashboard that hides model limitations can create automation bias, in which users accept a computer output without adequate scrutiny.

Hybrid modelling is often valuable in environmental engineering. Mechanistic mass balances, hydraulics and reaction kinetics can be combined with ML residual models or parameter estimation. This can improve extrapolation and physical plausibility, although it does not remove the need for validation.

6. Data Governance, Cybersecurity and Algorithmic Ethics

6.1 Data governance

Data governance assigns authority and accountability for data. It addresses ownership, stewardship, quality, metadata, access, retention, sharing, intellectual property, privacy and audit. Environmental data may include private-property information, worker locations, industrial process information or community-level exposure indicators. Open-data principles should be balanced with safety, privacy and commercial confidentiality.

A governance framework should define:

• a data catalogue and metadata standard;
• authoritative sources and version control;
• roles for owner, steward, analyst and approver;
• access based on least privilege;
• retention and disposal rules;
• procedures for correction and dispute;
• audit logs and reproducibility requirements; and
• incident reporting and recovery responsibilities.

6.2 Cybersecurity

Environmental infrastructure is critical infrastructure. Cybersecurity controls should address confidentiality, integrity and availability, with integrity and availability often especially important for treatment and control systems. Good practice includes asset inventories, network segmentation between information technology and operational technology, multifactor authentication, secure configuration, patch management, backups, vulnerability management, supplier assurance, tested incident response and offline recovery procedures.

Safety engineering and cybersecurity must be integrated. A cyberattack that changes a chemical dose is a process-safety event, not merely an information-technology problem. Remote access should be time-limited, logged and approved. Default passwords and unsupported devices should not be accepted in critical systems.

6.3 Algorithmic ethics

Ethical evaluation includes legality, beneficence, non-maleficence, justice, autonomy, transparency and accountability. A predictive maintenance system may prioritise affluent areas because they have denser sensor coverage, leaving poorly monitored communities at greater risk. A flood-risk model may reproduce historical underinvestment. Fairness therefore requires examining outcomes across locations and populations, not simply removing demographic variables.

Environmental decisions also raise intergenerational questions. Optimising current operating cost at the expense of long-term resource depletion is not sustainable. Governance should require impact assessments, stakeholder participation where decisions affect communities, appeal mechanisms and periodic independent review.

7. Advanced Membranes and Resource Recovery

7.1 Advanced membrane systems

Membrane processes separate constituents through a selective barrier. Microfiltration and ultrafiltration primarily remove suspended solids and microorganisms; nanofiltration and reverse osmosis provide greater removal of dissolved constituents. Forward osmosis, membrane distillation and membrane bioreactors offer specialised capabilities but may involve higher complexity or energy use.

Important design and operating concepts include flux, permeability, transmembrane pressure, concentration polarisation and fouling. A simplified relationship for a pressure-driven membrane is

J = A(ΔP − Δπ),

where J is solvent flux, A is permeability, ΔP is the transmembrane pressure difference and Δπ is the osmotic-pressure difference. This relation is an idealised form. It assumes a suitable solution, stable membrane properties and negligible or separately accounted fouling and concentration-polarisation effects. In real systems, fouling changes effective resistance and must be represented through calibration or more detailed resistance-in-series models.

Membrane selection must consider feed composition, recovery, concentrate management, cleaning chemicals, energy, pretreatment, lifespan and the fate of rejected contaminants. A membrane is not automatically sustainable: concentrate disposal and replacement materials may dominate impacts.

7.2 Resource recovery and circularity

Modern wastewater facilities can recover water, nutrients, energy, biopolymers and useful carbon products. Anaerobic digestion converts biodegradable organic matter into biogas; phosphorus can be recovered as struvite under suitable chemical conditions; nitrogen can be recovered or transformed through processes such as ammonia stripping or partial nitritation–anammox. Solid-waste systems can recover metals, plastics and organic fractions, while engineered systems may recover heat from wastewater.

Recovery should be evaluated by net benefit rather than gross output. The assessment should include energy and chemical inputs, contamination, market demand, transport, product quality, avoided-product credits and risks from concentrating hazardous substances. Life-cycle assessment and techno-economic analysis should be used together, because a process can have low greenhouse-gas emissions but be financially impractical, or be profitable while transferring pollution elsewhere.

8. Designing Future Net-Zero Environmental Systems

Net-zero environmental infrastructure requires whole-system accounting. Direct emissions, purchased energy, process emissions, fugitive methane and nitrous oxide, embodied carbon in construction materials, chemical production and end-of-life impacts may all be relevant. Offsets should not substitute for feasible reductions and should be reported separately from actual emission performance.

A credible transition pathway generally combines:

• demand management and leakage reduction;
• energy-efficient pumping, aeration and heating;
• renewable electricity and storage;
• methane capture and oxidation control;
• process optimisation and low-carbon chemicals;
• nutrient and material recovery;
• climate-resilient design; and
• transparent measurement, reporting and verification.

Digital twins and AI can support this transition by forecasting demand, coordinating distributed energy, detecting leaks and comparing retrofit options. They can also increase energy use through data centres, sensors and communications. Digital sustainability therefore requires measuring the environmental burden of digital infrastructure itself.

9. Illustrative Case: Digital Twin for a Municipal Wastewater Utility

Background

The following is a fictional illustrative case based on common utility conditions. A municipal wastewater utility operates an activated-sludge plant with variable inflow, high aeration energy use and occasional ammonia-limit excursions. Its sewer network also experiences wet-weather inflow and infiltration.

Problem

Operators rely on fixed aeration set points and periodic laboratory data. The plant has online dissolved-oxygen and ammonia sensors, but sensor fouling and inconsistent calibration reduce confidence. The utility proposes an AI-controlled digital twin.

Analysis

The engineering team first identifies the decision boundary: the system may recommend or adjust aeration set points, but it may not disable independent high-ammonia alarms or safety interlocks. A BIM–GIS asset register links tank geometry, blowers, air valves, sewer catchments and receiving-water constraints. Historical SCADA, laboratory, weather and maintenance data are reconciled using common time stamps and asset identifiers.

Data-quality analysis discovers that several ammonia readings during calibration were recorded as genuine process data. These records are flagged and excluded from model training. Validation is performed by holding out an entire wet-weather period rather than randomly selecting individual records. A mechanistic oxygen-demand model is combined with a forecasting model for inflow and ammonia load.

Approach and solution

The twin displays current states, estimates sensor reliability and forecasts dissolved oxygen and ammonia over a short horizon. An MPC layer proposes blower settings subject to minimum mixing, effluent ammonia limits, blower capacity and a conservative fallback mode. Operators approve initial recommendations. After a controlled trial, limited automatic adjustment is permitted during normal conditions, while unusual inputs trigger operator review.

Cybersecurity controls include network segmentation, multifactor authentication, signed software updates, local control capability and tested restoration from backups. A governance board approves model changes and reviews performance monthly. The utility reports energy, compliance, alarm frequency, false alarms, model drift and equity of service across catchments.

Outcome

In the illustrative case, aeration energy decreases during stable low-load periods, while compliance risk is reduced by earlier detection of rising ammonia. Benefits are not attributed solely to AI: sensor maintenance, data cleaning, operator training and process retuning are equally important. During a communications outage, local control continues safely and the system returns to a conservative set point.

Learning points

The case demonstrates that a digital twin is an organisational capability, not merely software. Process understanding, trustworthy data, constrained control, cybersecurity, human authority and continuous validation are necessary. The most valuable result may be improved operational discipline rather than a sophisticated algorithm.

10. Implementation and Professional Practice

A staged implementation reduces risk. Begin with a clearly bounded use case and baseline current performance. Establish data ownership and cybersecurity requirements before procurement. Pilot the system in shadow mode, compare it with existing practice, and define stop criteria. Train operators to challenge outputs and record overrides. After deployment, monitor drift, sensor health, model performance, energy use, environmental outcomes and unintended effects.

Procurement documents should require interoperable data formats, open interfaces where feasible, exportability, cybersecurity responsibilities, software-support periods, audit access and end-of-life arrangements. Vendors should not be permitted to make untestable claims about “AI optimisation.” Performance guarantees must specify baseline, operating conditions, measurement method and treatment of uncertainty.

Research and professional practice should increasingly combine field experiments, process modelling, socio-technical analysis and life-cycle assessment. Reproducibility requires documented data provenance, model versions, assumptions and decision logs. The environmental engineer’s central professional duty remains protection of public health and the environment; digital tools support that duty but do not replace judgement.

Challenges and Limitations

Digital environmental engineering faces persistent limitations:

• sensors drift, fail and may measure proxies rather than target variables;
• historical data can encode inequity and operational bias;
• rare extreme events are poorly represented in training data;
• models may fail when climate, land use or process conditions change;
• proprietary platforms can create vendor dependence;
• automation can weaken operator skills and situational awareness;
• cybersecurity upgrades may conflict with legacy equipment;
• advanced membranes can shift pollution into concentrates;
• resource-recovery products may contain contaminants or lack markets; and
• the energy and material footprint of digital systems can offset benefits.

These limitations favour transparent, modular and fail-safe designs. Simple robust rules may outperform complex AI when data are sparse or consequences are severe.

Contemporary Developments

Emerging directions include physics-informed machine learning, federated learning that keeps sensitive data at local sites, autonomous laboratories, hyperspectral remote sensing, low-power edge AI, digital product passports, carbon-aware computing and integrated urban digital twins. Generative AI may assist with document search, coding, inspection reports and scenario exploration, but generated outputs require expert verification and must not be treated as authoritative evidence without provenance.

Future systems will increasingly connect water, energy, waste, transport, buildings and ecological assets. Such integration can reveal synergies—for example, using recovered biogas and coordinating wastewater heat with district energy—but also creates coupled risks. Governance must therefore extend beyond individual facilities to regional infrastructure systems.

Chapter Summary

Digital environmental engineering links physical infrastructure with sensing, data, models, computation and controlled action. Digital models, shadows and twins differ in connectivity and operational capability; a true digital twin requires purpose, trusted data, validated models, uncertainty management and governance. BIM–GIS integration connects detailed facility information with geographic and environmental context across the asset lifecycle.

IoT systems require careful sensor selection, calibration, metadata, edge resilience and data-quality controls. Automation and optimisation must use explicit objectives and constraints, while MPC and AI should operate within safety boundaries. Machine learning is useful for forecasting, classification and anomaly detection, but requires representative data, leakage-free validation, uncertainty communication and human accountability.

Data governance, cybersecurity and algorithmic ethics are central engineering concerns. Advanced membranes and resource-recovery systems can support circularity, but their full energy, chemical, concentrate, contamination and market implications must be assessed. Net-zero systems require whole-life accounting and should combine efficiency, renewable energy, emission control, recovery and resilient operation. Digital tools are valuable when they improve measurable environmental outcomes without compromising safety, equity or professional responsibility.

Key Terms

Digital twin: A connected digital representation of a physical system used for monitoring, prediction, diagnosis or control.

Digital shadow: A digital representation updated by the physical system, generally without automatic feedback control.

BIM: Building information modelling; structured information about built assets and their lifecycle.

GIS: Geographic information system for managing and analysing spatial data.

IoT: Internet of Things; connected sensors and devices that collect and exchange data.

SCADA: Supervisory control and data acquisition system used to monitor and control industrial processes.

Edge computing: Processing data near the source rather than exclusively in a remote platform.

Data provenance: A record of where data originated, how they were transformed and who handled them.

Model predictive control: Receding-horizon control that forecasts process behaviour and optimises future actions.

Machine learning: Computational methods that infer patterns or relationships from data.

Data leakage: Unintended use of information in model development that would not be available at prediction time.

Algorithmic bias: Systematic distortion in automated outputs that produces unjust or inaccurate results.

Cyber-physical system: A system combining physical processes with computation, communication and control.

Advanced membrane: A selective separation barrier used for high-performance treatment or resource recovery.

Flux: Rate of fluid transfer through a membrane per unit membrane area.

Fouling: Accumulation of material that reduces membrane performance.

Resource recovery: Extraction of useful water, energy, nutrients or materials from waste streams.

Life-cycle assessment: Method for evaluating environmental impacts across a product or system’s life cycle.

Net zero: A condition in which relevant greenhouse-gas emissions are balanced by removals after feasible reductions.

Review Questions

Short Answer Questions

1. Distinguish a digital model, digital shadow and digital twin.

2. State four important dimensions of IoT data quality.

3. Why should environmental AI systems communicate uncertainty?

4. Identify three cybersecurity controls relevant to treatment-plant automation.

5. What factors must be considered when evaluating an advanced membrane system?

Long Answer Questions

1. Explain how BIM–GIS integration supports the planning, operation and renewal of environmental infrastructure.

2. Describe a responsible machine-learning workflow for predicting wastewater-treatment performance.

3. Discuss how constrained optimisation and model predictive control can reduce energy use without compromising compliance.

4. Evaluate the governance and ethical challenges of AI-assisted environmental decisions.

5. Explain how resource recovery and digitalisation can contribute to future net-zero environmental systems.

Analytical / Application Questions

1. A utility wants to automate aeration control using historical SCADA data. Develop a staged implementation plan that addresses data quality, validation, safety, cybersecurity and operator authority.

2. A flood-risk model performs well overall but systematically underpredicts risk in a low-income district with few sensors. Analyse possible causes and propose technical and governance remedies.

3. Compare two proposed membrane upgrades: one has lower energy demand but produces a difficult concentrate, while the other uses more energy and enables higher-quality water reuse. Design a decision framework using life-cycle, economic, operational and ethical criteria.

Multiple Choice Questions with Answers

Question 1

Which feature most clearly distinguishes a digital twin from a digital model?

A. It contains three-dimensional geometry

B. It is continuously connected to the physical system and can support operational decisions

C. It is stored in a cloud database

D. It uses artificial intelligence

Correct Answer: B. It is continuously connected to the physical system and can support operational decisions

Question 2

Which practice best prevents future information from entering a time-series model’s training data?

A. Randomly shuffling all observations

B. Using only the most recent observations

C. Using time-based validation

D. Removing missing values without recording the change

Correct Answer: C. Using time-based validation

Question 3

In the membrane relation J = A(ΔP − Δπ), J represents:

A. Solvent flux

B. Membrane area

C. Osmotic pressure

D. Chemical concentration

Correct Answer: A. Solvent flux

Question 4

Which control is most directly associated with limiting unauthorised access to operational systems?

A. Multifactor authentication

B. Increasing screen brightness

C. Reducing laboratory sampling

D. Changing pipe diameter

Correct Answer: A. Multifactor authentication

Question 5

Why can classification accuracy be misleading for rare environmental contamination events?

A. Accuracy has units of pressure

B. A model can appear accurate by predicting the common negative class

C. Accuracy cannot be calculated from observations

D. Rare events always have equal positive and negative cases

Correct Answer: B. A model can appear accurate by predicting the common negative class

Question 6

What is the principal purpose of a persistent asset identifier in BIM–GIS integration?

A. To eliminate the need for surveying

B. To link the same asset reliably across systems and its lifecycle

C. To increase sensor battery voltage

D. To replace all regulatory records

Correct Answer: B. To link the same asset reliably across systems and its lifecycle

Question 7

Which statement about model predictive control is correct?

A. It applies the entire forecast sequence without new measurements

B. It repeatedly forecasts, optimises and normally implements the first action

C. It never requires process constraints

D. It can operate safely without fallback control in every situation

Correct Answer: B. It repeatedly forecasts, optimises and normally implements the first action

Question 8

Which is an example of algorithmic fairness concern in environmental engineering?

A. A model uses a documented unit conversion

B. A maintenance model prioritises areas with dense sensors and neglects poorly monitored communities

C. A sensor is calibrated against a reference method

D. A pump has a defined operating limit

Correct Answer: B. A maintenance model prioritises areas with dense sensors and neglects poorly monitored communities

Question 9

Which statement best describes resource recovery?

A. It means disposing of waste at the lowest immediate cost

B. It extracts useful water, energy, nutrients or materials while considering whole-system impacts

C. It always eliminates all hazardous substances

D. It applies only to municipal solid waste

Correct Answer: B. It extracts useful water, energy, nutrients or materials while considering whole-system impacts

Question 10

Why should digital tools be included in sustainability assessment?

A. They have no material or energy footprint

B. Their sensors, communications and computing can create environmental impacts

C. They automatically guarantee net-zero operation

D. They remove the need for life-cycle assessment

Correct Answer: B. Their sensors, communications and computing can create environmental impacts