Water-System Risk, Extreme Droughts, Flood Infrastructure and Reliability

Unit 07 — Water-System Risk

Unit overview. Risk combines event probability with exposure, consequence, preparedness and recovery. Return period is often misread as a timetable: an event with annual exceedance probability 1/T can recur in consecutive years. Drought evolves differently across meteorological precipitation, agricultural soil moisture and hydrological storage; indicators have different lead times. Climate variability and nonstationarity challenge methods built solely on historical distributions. Water infrastructure can fail through pumps, power supply, cybersecurity, contamination, structural damage and supply-chain interruption even when the river itself has adequate flow. Robust planning asks whether an acceptable service can be maintained across plausible scenarios.

Field problem. Compare a town supplied by one large pumping station with a town using several smaller sources. On average the single station may have lower unit cost, but a flood may disconnect its power and road access simultaneously. Test both economic efficiency and failure-mode diversity. Describe a warning threshold that is actionable, not merely statistically striking: how much lead time is needed to deliver water, switch intakes or issue drought restrictions?

Four topics. Drought, Floods, Climate Variability, Infrastructure Reliability.

Before starting: Sketch a control volume and record which terms describe depths, rates, stored volumes, concentrations, entitlements or monetary values. Identify the boundary, the timescale, the user groups affected and at least one observation whose absence would make a numerical conclusion unreliable. Explain why seasonal variability matters more than a single annual figure for some decisions.

25. Drought

Core concept. Drought indicators identify precipitation shortage, soil-water deficits or depleted hydrological storage. A standardized anomaly is a simple educational screen; it is not interchangeable with the statistically fitted Standardized Precipitation Index.

Analysis workflow. Compute a z-score relative to a synthetic historic mean and standard deviation, then test sensitivity to the sample window. Compare a warning driven by precipitation with one driven by actual storage.

Engineering interpretation. Early warning is useful only if restrictions, water trucking or emergency supplies can be implemented in time.

Evidence and measurement. Work with an illustrative upland watershed where the available record includes rainfall and gauge coverage. Begin by naming the controlled variable, the response variable and the geographic location of each observation. Check the date, measurement frequency, applicable units and any uncertainty statement. Where a record is missing or derived from a different location, describe the interpolation explicitly rather than silently assuming that it is representative. Before using a numerical method, distinguish a conservation equation from an empirical relation and a legally mandated constraint. Only conservation identities hold without calibration.

From conservation to prediction. A transparent first calculation should reconcile a mass, energy or financial balance as appropriate to the physical mechanism. Convert input quantities into coherent dimensions and display intermediate steps, particularly when mixing depth, area, volume, discharge and time. If a relationship uses a calibrated coefficient, record the period and conditions that support it. Run at least two alternative parameter sets, recording how the predicted outcome changes. A sensitivity experiment identifies leverage but does not by itself assign a probability to uncertainty.

Scenario comparison. Compare a baseline in which current practice continues against an alternative with a different capacity, demand, governance rule or environmental reserve. Where possible, change only one assumption at a time, then carry out a second experiment in which multiple uncertain inputs vary together. Explain what each comparison isolates and what interactions remain hidden. Keep the original measurements, assumptions, units and script version in a reproducible notebook. A repeatable result is necessary but insufficient evidence of a valid model.

Scope and limitations. Do not interpret a screening model as an approved infrastructure design. Confirm whether physical boundary conditions, environmental constraints, public-health requirements and downstream effects appear in the calculation. A conceptual model that omits return flows, aquifer exchange, water quality or operational failures may overestimate usable supply. Sketch a conceptual system with each inflow, outflow, storage and feedback represented. A number without a transparent system boundary is not an auditable engineering result.

Practice questions with answer guidance.

1. Explain: How does the concept of drought connect to both the quantity and timing of water availability in water-system risk? Answer guidance: Identify the physically relevant boundary, state the controlling process described above and name one observation that could falsify the assumed behavior.
2. Quantify: Choose two contrasting data periods for the upland watershed, state all units and compute one meaningful normalized comparison; distinguish an absolute difference from a percentage change. Answer guidance: Report the numerator, denominator, observation period, units and rounding precision; do not manufacture a statistical confidence interval without an error model.
3. Diagnose: Suppose a simple model fits the first half of the observation period but fails after a large storm or a prolonged dry season. Which neglected process would you investigate first? Answer guidance: Look for a process whose magnitude or regime changes with the event, then specify a measurement that could discriminate it from an input error.
4. Decide: What evidence would justify changing a withdrawal rule, reservoir policy or monitoring plan? Answer guidance: Compare safety and minimum-service constraints first; then evaluate operating performance, maintenance cost, affected stakeholders and the reversibility of the decision.

Mini field task. Write a one-page technical memo on drought for the upland watershed. Include a schematic boundary, a data-quality table, one auditable calculation, a counter-scenario and a clear statement of what the evidence does not establish. Do not present synthetic numbers as observed field data.

26. Floods

Core concept. Flood risk depends on exceedance probability, inundation depth, exposed assets and vulnerability. The expected count of exceedances over time is not a guarantee, and consecutive large floods are possible.

Analysis workflow. Compute the probability of at least one event over a design lifetime under an explicitly assumed stationary independent annual-event model. Discuss how flood protection changes consequence without eliminating probability.

Engineering interpretation. Dam breach, compound floods, urban drainage blockage and changing exposure may dominate a simple probability calculation.

Evidence and measurement. Work with an illustrative seasonal monsoon basin where the available record includes measured flow and stage. Begin by naming the controlled variable, the response variable and the geographic location of each observation. Check the date, measurement frequency, applicable units and any uncertainty statement. Where a record is missing or derived from a different location, describe the interpolation explicitly rather than silently assuming that it is representative. Separate a model output from an observation. A discrepancy may reveal a missing process, incorrectly measured inputs, model-structure error or a unit conversion mistake.

From conservation to prediction. A transparent first calculation should reconcile a mass, energy or financial balance as appropriate to the physical mechanism. Convert input quantities into coherent dimensions and display intermediate steps, particularly when mixing depth, area, volume, discharge and time. If a relationship uses a calibrated coefficient, record the period and conditions that support it. Specify an observable metric and the unit in which it would be reported. A useful metric permits comparisons across dates without changing the spatial boundary or its denominator.

Scenario comparison. Compare a baseline in which current practice continues against an alternative with a different capacity, demand, governance rule or environmental reserve. Where possible, change only one assumption at a time, then carry out a second experiment in which multiple uncertain inputs vary together. Explain what each comparison isolates and what interactions remain hidden. Before using a numerical method, distinguish a conservation equation from an empirical relation and a legally mandated constraint. Only conservation identities hold without calibration.

Scope and limitations. Do not interpret a screening model as an approved infrastructure design. Confirm whether physical boundary conditions, environmental constraints, public-health requirements and downstream effects appear in the calculation. A conceptual model that omits return flows, aquifer exchange, water quality or operational failures may overestimate usable supply. Run at least two alternative parameter sets, recording how the predicted outcome changes. A sensitivity experiment identifies leverage but does not by itself assign a probability to uncertainty.

Practice questions with answer guidance.

1. Explain: How does the concept of floods connect to both the quantity and timing of water availability in water-system risk? Answer guidance: Identify the physically relevant boundary, state the controlling process described above and name one observation that could falsify the assumed behavior.
2. Quantify: Choose two contrasting data periods for the seasonal monsoon basin, state all units and compute one meaningful normalized comparison; distinguish an absolute difference from a percentage change. Answer guidance: Report the numerator, denominator, observation period, units and rounding precision; do not manufacture a statistical confidence interval without an error model.
3. Diagnose: Suppose a simple model fits the first half of the observation period but fails after a large storm or a prolonged dry season. Which neglected process would you investigate first? Answer guidance: Look for a process whose magnitude or regime changes with the event, then specify a measurement that could discriminate it from an input error.
4. Decide: What evidence would justify changing a withdrawal rule, reservoir policy or monitoring plan? Answer guidance: Compare safety and minimum-service constraints first; then evaluate operating performance, maintenance cost, affected stakeholders and the reversibility of the decision.

Mini field task. Write a one-page technical memo on floods for the seasonal monsoon basin. Include a schematic boundary, a data-quality table, one auditable calculation, a counter-scenario and a clear statement of what the evidence does not establish. Do not present synthetic numbers as observed field data.

27. Climate Variability

Core concept. Natural modes and long-term climate change can shift rainfall seasonality, evapotranspiration, runoff and extremes. Planning should evaluate physically plausible stress tests in addition to historical extrapolation.

Analysis workflow. Perturb inflow means, storm concentration and demand separately; compare design ranking across cases without assigning unsupported future probabilities.

Engineering interpretation. A large ensemble can still miss a critical failure mode if every scenario shares the same structural assumption.

Evidence and measurement. Work with an illustrative peri-urban supply district where the available record includes head and storage records. Begin by naming the controlled variable, the response variable and the geographic location of each observation. Check the date, measurement frequency, applicable units and any uncertainty statement. Where a record is missing or derived from a different location, describe the interpolation explicitly rather than silently assuming that it is representative. Sketch a conceptual system with each inflow, outflow, storage and feedback represented. A number without a transparent system boundary is not an auditable engineering result.

From conservation to prediction. A transparent first calculation should reconcile a mass, energy or financial balance as appropriate to the physical mechanism. Convert input quantities into coherent dimensions and display intermediate steps, particularly when mixing depth, area, volume, discharge and time. If a relationship uses a calibrated coefficient, record the period and conditions that support it. Choose a realistic time step. Storm peaks may need minutes while aquifer or water-allocation trends may be represented over months; coarse averaging can erase the limiting event.

Scenario comparison. Compare a baseline in which current practice continues against an alternative with a different capacity, demand, governance rule or environmental reserve. Where possible, change only one assumption at a time, then carry out a second experiment in which multiple uncertain inputs vary together. Explain what each comparison isolates and what interactions remain hidden. Separate a model output from an observation. A discrepancy may reveal a missing process, incorrectly measured inputs, model-structure error or a unit conversion mistake.

Scope and limitations. Do not interpret a screening model as an approved infrastructure design. Confirm whether physical boundary conditions, environmental constraints, public-health requirements and downstream effects appear in the calculation. A conceptual model that omits return flows, aquifer exchange, water quality or operational failures may overestimate usable supply. Specify an observable metric and the unit in which it would be reported. A useful metric permits comparisons across dates without changing the spatial boundary or its denominator.

Practice questions with answer guidance.

1. Explain: How does the concept of climate variability connect to both the quantity and timing of water availability in water-system risk? Answer guidance: Identify the physically relevant boundary, state the controlling process described above and name one observation that could falsify the assumed behavior.
2. Quantify: Choose two contrasting data periods for the peri-urban supply district, state all units and compute one meaningful normalized comparison; distinguish an absolute difference from a percentage change. Answer guidance: Report the numerator, denominator, observation period, units and rounding precision; do not manufacture a statistical confidence interval without an error model.
3. Diagnose: Suppose a simple model fits the first half of the observation period but fails after a large storm or a prolonged dry season. Which neglected process would you investigate first? Answer guidance: Look for a process whose magnitude or regime changes with the event, then specify a measurement that could discriminate it from an input error.
4. Decide: What evidence would justify changing a withdrawal rule, reservoir policy or monitoring plan? Answer guidance: Compare safety and minimum-service constraints first; then evaluate operating performance, maintenance cost, affected stakeholders and the reversibility of the decision.

Mini field task. Write a one-page technical memo on climate variability for the peri-urban supply district. Include a schematic boundary, a data-quality table, one auditable calculation, a counter-scenario and a clear statement of what the evidence does not establish. Do not present synthetic numbers as observed field data.

28. Infrastructure Reliability

Core concept. Reliable supply requires sources, treatment, energy, pumps, valves, communications, staff and storage to function together. Common-cause failures create dependence that is missed when each component is modeled as perfectly independent.

Analysis workflow. Construct a simple parallel-source availability model and compare it with a common-failure factor. Describe which neglected correlations could make predicted service optimistic.

Engineering interpretation. Maintenance, backup power and spare parts often yield better resilience per unit cost than expanding nominal raw-water capacity.

Evidence and measurement. Work with an illustrative river-delta corridor where the available record includes population and use records. Begin by naming the controlled variable, the response variable and the geographic location of each observation. Check the date, measurement frequency, applicable units and any uncertainty statement. Where a record is missing or derived from a different location, describe the interpolation explicitly rather than silently assuming that it is representative. Run at least two alternative parameter sets, recording how the predicted outcome changes. A sensitivity experiment identifies leverage but does not by itself assign a probability to uncertainty.

From conservation to prediction. A transparent first calculation should reconcile a mass, energy or financial balance as appropriate to the physical mechanism. Convert input quantities into coherent dimensions and display intermediate steps, particularly when mixing depth, area, volume, discharge and time. If a relationship uses a calibrated coefficient, record the period and conditions that support it. Keep the original measurements, assumptions, units and script version in a reproducible notebook. A repeatable result is necessary but insufficient evidence of a valid model.

Scenario comparison. Compare a baseline in which current practice continues against an alternative with a different capacity, demand, governance rule or environmental reserve. Where possible, change only one assumption at a time, then carry out a second experiment in which multiple uncertain inputs vary together. Explain what each comparison isolates and what interactions remain hidden. Sketch a conceptual system with each inflow, outflow, storage and feedback represented. A number without a transparent system boundary is not an auditable engineering result.

Scope and limitations. Do not interpret a screening model as an approved infrastructure design. Confirm whether physical boundary conditions, environmental constraints, public-health requirements and downstream effects appear in the calculation. A conceptual model that omits return flows, aquifer exchange, water quality or operational failures may overestimate usable supply. Choose a realistic time step. Storm peaks may need minutes while aquifer or water-allocation trends may be represented over months; coarse averaging can erase the limiting event.

Practice questions with answer guidance.

1. Explain: How does the concept of infrastructure reliability connect to both the quantity and timing of water availability in water-system risk? Answer guidance: Identify the physically relevant boundary, state the controlling process described above and name one observation that could falsify the assumed behavior.
2. Quantify: Choose two contrasting data periods for the river-delta corridor, state all units and compute one meaningful normalized comparison; distinguish an absolute difference from a percentage change. Answer guidance: Report the numerator, denominator, observation period, units and rounding precision; do not manufacture a statistical confidence interval without an error model.
3. Diagnose: Suppose a simple model fits the first half of the observation period but fails after a large storm or a prolonged dry season. Which neglected process would you investigate first? Answer guidance: Look for a process whose magnitude or regime changes with the event, then specify a measurement that could discriminate it from an input error.
4. Decide: What evidence would justify changing a withdrawal rule, reservoir policy or monitoring plan? Answer guidance: Compare safety and minimum-service constraints first; then evaluate operating performance, maintenance cost, affected stakeholders and the reversibility of the decision.

Mini field task. Write a one-page technical memo on infrastructure reliability for the river-delta corridor. Include a schematic boundary, a data-quality table, one auditable calculation, a counter-scenario and a clear statement of what the evidence does not establish. Do not present synthetic numbers as observed field data.

Unit application and simulation protocol

Laboratory 13 — Drought Anomaly Screen

Purpose and physical meaning. This laboratory isolates one process within water-resources engineering. It is intentionally a limited analytical or bookkeeping model rather than a digital twin of any real basin. The objective is to predict how an output responds to controlled input changes, reproduce the calculation independently and recognize the boundary beyond which the model loses credibility. Use the graph to explore variation, not as evidence of real-world accuracy.

Equation and units. z = (Pobs−μ)/σ. Inputs are observed rainfall in mm/year, reference mean in mm/year, and reference sd in mm/year. Units matter because a factor of 1,000 can arise when litres are confused with cubic metres, or when a daily rate is used directly in a monthly balance. Write out the conversion before comparing scenarios.

Default experiment. Start with the provided values: Observed rainfall = 680 mm/year, Reference mean = 950 mm/year, Reference SD = 170 mm/year. Write a verbal prediction of the first output. Make a hand calculation for the default setting wherever the equation permits it; for a time series, reproduce the first step and check that the total storage change equals the accumulation of fluxes. Only then compare with the displayed model output.

Sensitivity investigation. Move the first slider down to 250 mm/year and up to 1800 mm/year, holding the other two at their defaults. Examine the direction and shape of the response, and inspect whether the graph is discontinuous at a physical constraint such as zero storage or an allocation cap. Repeat for the second and third inputs. Export the curve to CSV and estimate a finite-difference sensitivity from adjacent rows; specify the unit of the derivative.

Boundary and validation. A standardized rainfall anomaly is not a fitted Standardized Precipitation Index. Changes in reference period and rainfall skewness affect classification. A credible validation task would compare the predicted process with a data record measured under comparable conditions and quantify deviations against uncertainties. Do not tune an input to force agreement and then claim independent verification using the same observations. Distinguish a passing arithmetic check from a validated hydrological model.

Challenge. How would changing the baseline period alter drought labeling even with the same observed rainfall? Report how the answer changes under an extreme but physically plausible input combination, and identify one omitted mechanism that would be important before real implementation. In your conclusion distinguish observations, model assumptions and proposed decisions. If the simulation becomes insensitive to an input, determine whether a maximum/minimum constraint has become active or an equation is coded incorrectly.

Student deliverables. Include a hand calculation, a three-row baseline/low/high scenario table with units, the exported response plot, a one-paragraph discussion of parameter sensitivity, and a one-paragraph critique of model applicability. Name the specific monitoring records and institutional approvals a responsible engineer would seek before designing a project.

Laboratory 14 — Flood Lifetime Exceedance

Purpose and physical meaning. This laboratory isolates one process within water-resources engineering. It is intentionally a limited analytical or bookkeeping model rather than a digital twin of any real basin. The objective is to predict how an output responds to controlled input changes, reproduce the calculation independently and recognize the boundary beyond which the model loses credibility. Use the graph to explore variation, not as evidence of real-world accuracy.

Equation and units. P(at least one) = 1 − (1−1/T)^N. Inputs are return period in years, exposure horizon in years, and asset value in million currency units. Units matter because a factor of 1,000 can arise when litres are confused with cubic metres, or when a daily rate is used directly in a monthly balance. Write out the conversion before comparing scenarios.

Default experiment. Start with the provided values: Return period = 100 years, Exposure horizon = 30 years, Asset value = 150 million currency units. Write a verbal prediction of the first output. Make a hand calculation for the default setting wherever the equation permits it; for a time series, reproduce the first step and check that the total storage change equals the accumulation of fluxes. Only then compare with the displayed model output.

Sensitivity investigation. Move the first slider down to 5 years and up to 500 years, holding the other two at their defaults. Examine the direction and shape of the response, and inspect whether the graph is discontinuous at a physical constraint such as zero storage or an allocation cap. Repeat for the second and third inputs. Export the curve to CSV and estimate a finite-difference sensitivity from adjacent rows; specify the unit of the derivative.

Boundary and validation. Assumes independent annual events and stationary exceedance probability. Asset exposure is illustrative; probability is not expected damage. A credible validation task would compare the predicted process with a data record measured under comparable conditions and quantify deviations against uncertainties. Do not tune an input to force agreement and then claim independent verification using the same observations. Distinguish a passing arithmetic check from a validated hydrological model.

Challenge. Can a 100-year event happen twice in ten years? What breaks the independence or stationarity assumption? Report how the answer changes under an extreme but physically plausible input combination, and identify one omitted mechanism that would be important before real implementation. In your conclusion distinguish observations, model assumptions and proposed decisions. If the simulation becomes insensitive to an input, determine whether a maximum/minimum constraint has become active or an equation is coded incorrectly.

Student deliverables. Include a hand calculation, a three-row baseline/low/high scenario table with units, the exported response plot, a one-paragraph discussion of parameter sensitivity, and a one-paragraph critique of model applicability. Name the specific monitoring records and institutional approvals a responsible engineer would seek before designing a project.

Applied project

Project 07 — Water-System Risk: basin decision memorandum

Decision setting. Compare a town supplied by one large pumping station with a town using several smaller sources. On average the single station may have lower unit cost, but a flood may disconnect its power and road access simultaneously. Test both economic efficiency and failure-mode diversity. Describe a warning threshold that is actionable, not merely statistically striking: how much lead time is needed to deliver water, switch intakes or issue drought restrictions?

Project mandate. Assume a local water commission must make a reversible decision within a limited financial and operational envelope. The project is not to design a real structure; it is to demonstrate reproducible calculations and show when the engineering team must request additional information. Assign a stakeholder team to challenge the baseline scenario and a second team to challenge the preferred intervention. Students should explicitly document disagreement rather than forcing the same preference through all models.

Required tasks.

1. Define the physical system boundary and provide a map or schematic of inflow, outflow, demands and stores.
2. Write an auditable water balance and an explicit list of observations needed to populate it.
3. Compare a baseline, a dry-period stress case and a governance or engineering intervention.
4. Quantify a service-reliability indicator together with one ecological or distributional consequence.
5. State at least two risks not represented by the introductory classroom models.
6. Propose a monitoring plan with observations, frequency, responsible parties and decision triggers.
7. Prepare a transparent summary for people affected upstream and downstream, separating confirmed evidence from assumptions.

Method. Investigate all four syllabus topics of Unit 7: Drought, Floods, Climate Variability, Infrastructure Reliability. Use the associated pair of simulations as teaching tools, independently verify at least three quantitative results, and do not insert those simulation outputs directly into a permit or construction drawing. When two assumptions point to opposite decisions, describe the range of plausible outcomes and identify which field measurement would most cheaply reduce uncertainty. An optimized number without enforceable operational responsibilities is not a complete solution.

Submission rubric. Technical closure (25%), evidence quality and assumptions (20%), scenario sensitivity and uncertainty (20%), environmental and equity impacts (15%), reproducibility of calculations and charts (10%), and clarity of recommendations (10%). A polished graph cannot compensate for an incorrectly defined basin boundary or missing feasibility constraint.

Unit 07 synthesis and review

As a final check, reconcile the data and assumptions across the four topics rather than treating them as isolated calculations. Prepare a source table showing which information is a measurement, which comes from a model and which expresses a policy decision. Compare what would happen at daily, seasonal and multiannual time scales. Identify one conflict between upstream and downstream objectives and one environmental consequence of the most attractive numerical alternative. Record the monitoring action that would most improve the evidence base. A good water-resources engineer must know what the calculations leave unresolved.

---