Choose your reading route
eduKateSG · Why Mathematics? · Resilience Metrics: Performance Loss, Recovery Curves, Time to Recover and System Capability
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 11–20
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 21–30
Check assumptions
See where design and calibration matter.Route 4 · Chapters 31–40
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 41–47
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Why is mathematics important in resilience metrics? Reliability asks whether a function fails; resilience also asks how much performance is lost, how long disruption lasts and how recovery unfolds. A resilience curve turns service level over time into magnitudes, durations, thresholds and integrals that can be compared transparently.
This guide develops performance loss, recovery time, area-based measures and uncertainty with practical examples. The numbers are educational. Real community or infrastructure decisions require stakeholder-defined functions, hazard scenarios, distributional impacts and competent governance.
Full chapter index
Chapters 1–10 · Understand the question
Chapters 11–20 · Reproduce the example
Chapters 21–30 · Check assumptions
Chapters 31–40 · Explore applications
- Validation checks
- Student activity
- Parent guidance
- Common misconception: faster recovery always means greater resilience
- Common misconception: one index is universal
- Did You Know?
- Practice questions
- FAQ: is resilience just reliability plus repair?
- FAQ: should we use the resilience triangle?
- Reporting checklist
Chapters 41–47 · Study and verify
Reliability often concerns avoiding failure over a mission. Resilience includes absorption, degraded operation, adaptation and recovery after disruption. A highly reliable system can recover slowly after a rare failure; a less reliable one may recover quickly. State which concept and horizon the metric represents.
Let Q(t) measure delivered function, such as proportion of demand served, safe throughput or accessible capacity. Normalise only when the reference level is meaningful. A single technical output may miss social consequences, quality or unequal access. Stakeholders should help define performance.
Choose pre-event baseline Q_0, minimum acceptable target and desired recovery level. Full restoration may be impossible or unnecessary. A changing baseline complicates comparison. Record season, demand and population so a percentage has a denominator.
Suppose normalised performance is 1 before an event, drops instantly to 0.6, remains there for 2 days and then recovers linearly to 1 over 8 more days. The maximum loss is 0.4 and time to full recovery is 10 days.
The loss area integrates 1-Q(t) over the control interval. The 2-day plateau contributes 0.4×2=0.8 performance-days. Linear recovery contributes a triangle 0.5×8×0.4=1.6. Total loss area is 2.4 performance-days.
Over a 10-day interval, average performance is 1-2.4/10=0.76. This area-based index combines loss and duration. It depends on interval and reference. Extending the interval after full recovery raises the index, so control windows must be comparable.
| Metric | Value |
|---|---|
| Initial performance loss | 0.40 |
| Plateau loss area | 0.80 performance-days |
| Recovery loss area | 1.60 performance-days |
| Total loss area | 2.40 performance-days |
| Ten-day area index | 0.76 |
The drop from 1 to 0.6 is 0.4, or 40% of baseline. Magnitude is easy to communicate but ignores duration and recovery. Two systems can have the same initial loss and very different service histories.
Recovery time can mean return to baseline, a target such as 90%, or a stable level for a defined duration. State the threshold and persistence rule. Temporary crossing should not count as completion if performance falls again.
A simple average recovery rate is performance regained divided by recovery duration: 0.4/8=0.05 per day in the example. Nonlinear recovery needs local slopes or phase-specific rates. Average rate can hide delays and reversals.
| Family | Question |
|---|---|
| Magnitude | How deep was the loss? |
| Duration | How long did disruption last? |
| Rate | How quickly did performance change? |
| Threshold | When was critical service restored? |
| Integral | How much service was lost over time? |
The classic triangle assumes an abrupt loss and roughly linear recovery. Real trajectories can cascade, plateau, oscillate, overshoot or never fully recover. Use the triangle as an illustration, not a universal shape. The curve taxonomy should match observed mechanisms.
A resilience trapezoid separates deterioration, degraded plateau, recovery and new steady state. Milestones help diagnose whether protection, response or restoration dominates. However, milestone definitions can be ambiguous and should be validated across trajectories.
Count hours below a critical service threshold or time until minimum service is restored. Threshold metrics can align with policy goals but may ignore improvements above the threshold. Report the full curve or complementary area measure.
Communities depend on power, water, transport, communications and health services. Combining them requires weights, substitution rules and dependency models. A weighted average can hide complete loss of one essential service. Report component curves as well as any composite.
Power failure can disable communications, which slows restoration, creating feedback. Static independent curves miss cascades. Use network, system-dynamics, agent or probabilistic models where dependencies materially affect recovery. Validate causal pathways.
Hazard intensity, damage, resource availability and repair duration are uncertain. Simulate an ensemble of trajectories and calculate each trajectory’s metrics. Summarise the distribution of metrics; do not compute only one metric from the mean trajectory because nonlinear operations differ.
A resilience number must name the hazard scenario, severity, timing and affected region. Comparing systems under different hazards is misleading. Use a scenario set with probabilities only when those probabilities are defensible.
Average performance loss can hide rare catastrophic trajectories. Report quantiles, exceedance probabilities or conditional tail measures where consequences justify them. Avoid reducing all uncertainty to one expected curve.
An average recovery can improve while vulnerable groups remain without service. Stratify curves by location or population where ethical and data governance permit. Weights reflect values and should be explicit, not buried in a composite index.
The availability guide measures proportion of time operational. Resilience curves preserve severity and recovery path after disruption. A binary availability trace is one special case of a performance curve.
The multi-state reliability guide treats discrete performance levels. Resilience adds temporal degradation and restoration. The same state values can become Q(t) along a trajectory.
The network-reliability guide asks whether terminals remain connected. Resilience can track fraction of demand connected, travel time or capacity during repair, not just final connectivity.
Repair order, crews, travel, spares and access constraints shape the curve. Optimisation can compare schedules, but objectives should include safety and equity as relevant. A faster average restoration may not serve the highest-consequence locations first.
Mitigation can reduce initial loss; redundant capacity can maintain a higher plateau; stockpiles and crews can shorten recovery. Resilience metrics show which phase changes. Economic analysis should include costs, uncertainty and consequences beyond the curve.
Robustness usually means limiting performance loss under disturbance, while rapidity concerns recovery speed. Redundancy and resourcefulness are related dimensions. Definitions vary across fields, so define terms rather than assuming one universal taxonomy.
Recovery may end above or below the original baseline. A new steady state above one can reflect improvement or a changed denominator. Separate restoration from long-term adaptation. Do not credit unrelated growth to the disruption response.
Record event time, baseline demand, performance observations, restoration milestones, resource deployments and measurement gaps. High-frequency technical data may still miss user experience. Preserve raw series and revisions so metrics can be reproduced.
Observation may end before full recovery. Time-to-recovery is then right censored, not equal to study length. Survival methods can estimate recovery distributions. Area loss over the observed window remains partial and must be labelled.
Sensors often fail during severe events. Interpolation can understate abrupt loss or delayed recovery. Flag missing intervals, use bounds or models with uncertainty and compare independent sources. A smooth line should not imply measured continuity.
Pre-event performance varies by day and season. Estimate a counterfactual baseline from comparable periods or models, then propagate uncertainty. A fixed baseline chosen after the event can bias loss area.
Recalculate the 2.4 performance-day loss, 0.76 index and 0.05-per-day recovery rate. Check units, interval and threshold definitions. For simulated curves, verify bounds on Q(t) and metric code using hand-built shapes.
Compare modelled and observed loss magnitude, plateau, milestones and recovery distribution in held-out events or exercises. Validate service meaning with operators and users. A curve can fit physical output while missing actual access.
Plot the example curve on graph paper, shade the loss area and calculate rectangle plus triangle. Change the plateau from 2 to 4 days and recompute. Explain which intervention would change magnitude versus recovery.
Ask the student what “back to normal” means and for whom. Encourage labelled axes, units and a shaded area before formulas. Compare a deep short disruption with a shallow long one to show why one metric is insufficient.
CHAPTER 34 OF 47 · Explore applications
34. Common misconception: faster recovery always means greater resilience
A system can recover quickly after a very deep loss yet deliver less total service than one with a small loss and slower return. Compare magnitude, duration, area and critical thresholds together.
CHAPTER 35 OF 47 · Explore applications
35. Common misconception: one index is universal
Index values depend on performance definition, normalisation, control interval and weights. Use a small dashboard of meaningful measures and the underlying curve. Avoid rankings that hide assumptions.
Taking the metric of an average trajectory can give a different result from averaging the metric across trajectories. Nonlinearity means uncertainty should be propagated through the calculation, not collapsed first.
Define Q(t), baseline and control interval. Reproduce the 2.4 loss area and 0.76 index. Distinguish robustness and rapidity. Explain censoring and equity. Design a metric for service below a critical threshold.
CHAPTER 38 OF 47 · Explore applications
38. FAQ: is resilience just reliability plus repair?
It is broader in many fields because it can include degraded performance, adaptation, multiple services and social outcomes. Definitions vary, so state scope. Reliability and maintainability remain important inputs.
CHAPTER 39 OF 47 · Explore applications
39. FAQ: should we use the resilience triangle?
Use it when the trajectory assumptions are adequate or as a teaching approximation. For cascading, plateaued, oscillating or incomplete recovery, represent the observed shape and select metrics accordingly.
Report hazard scenario, performance function, baseline, target, affected population, control interval, curve data, milestones, magnitude, duration, area, thresholds, uncertainty ensemble, censoring, equity, dependencies, validation and decision use.
NIST describes community resilience performance goals and assessment metrics and models for the economics of resilience. A review examines resilience curves and summary metrics.
Mathematics matters because resilience is a path through time, not a single uptime number. Curves, areas, thresholds and recovery distributions reveal different strengths and weaknesses. Responsible metrics define performance with stakeholders, propagate uncertainty and keep assumptions visible.
Standardise performance definitions and control intervals before comparing events. Severity adjustment may be needed, but modelled adjustment adds uncertainty. Report raw and adjusted metrics so readers can see how rankings change.
Preparedness inventories and exercise completion are leading indicators; observed service loss and recovery are lagging outcomes. Do not combine them into one score without a theory linking preparation to performance. Validate leading indicators against later results.
Publish milestone definitions, current uncertainty and revised estimates. A forecasted restoration time is not a promise. Update with new evidence and preserve previous forecasts for calibration and accountability.
Resilience analysis connects engineering, planning, economics, public policy and data science. Students use functions, integration, statistics and optimisation. Responsible work also needs community engagement, domain knowledge and ethical distribution of resources.
A student should draw a performance curve, compute magnitude, recovery time and loss area, then explain why an average index can hide thresholds and unequal impacts. A strong answer includes uncertainty and a clearly defined service.
Extended practice laboratory
Practice case 1: Initial loss
Setup: Drop Q from 1 to 0.6.
Task: Calculate magnitude.
Checking target: Obtain 0.4 or 40%.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 2: Plateau loss
Setup: Hold Q=0.6 for two days.
Task: Calculate rectangle area.
Checking target: Obtain 0.8 performance-days.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 3: Recovery loss
Setup: Recover linearly over eight days.
Task: Calculate triangle area.
Checking target: Obtain 1.6 performance-days.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 4: Total loss
Setup: Add both areas.
Task: Calculate cumulative shortfall.
Checking target: Obtain 2.4 performance-days.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 5: Area index
Setup: Use a ten-day interval.
Task: Calculate 1-2.4/10.
Checking target: Obtain 0.76.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 6: Recovery rate
Setup: Regain 0.4 over eight days.
Task: Calculate average slope.
Checking target: Obtain 0.05 per day.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 7: Threshold time
Setup: Choose Q=0.9.
Task: Find first persistent crossing.
Checking target: State the persistence rule.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 8: Nonlinear curve
Setup: Add a plateau and setback.
Task: Compare with a triangle.
Checking target: Do not force linear recovery.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 9: Trajectory ensemble
Setup: Simulate many curves.
Task: Compute each metric separately.
Checking target: Average metrics, not only the mean curve.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 10: Censoring
Setup: End observation before recovery.
Task: Label time to recovery censored.
Checking target: Do not equal it to study length.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 11: Equity
Setup: Split Q by population group.
Task: Compare curves and thresholds.
Checking target: Do not hide complete loss in an average.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.
Practice case 12: Conclusion
Setup: State hazard, service, baseline and interval.
Task: Report magnitude, duration, area and uncertainty.
Checking target: Separate forecast from promise.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows, identify the assumption that carries the most consequence, and describe one verification or validation check. Finish with a cautious sentence that distinguishes the illustrative calculation from a real operational approval.
Method note: For each case, sketch Q(t) with labelled units, baseline, milestones and control interval. Shade the loss area and identify whether the metric measures magnitude, duration, rate, threshold or integral. Then state whose service is represented, how uncertainty or censoring enters and what curve shape would make the simplified calculation misleading.
Reflection: Explain why the checking target follows from the setup, what evidence would be needed outside this illustration, and how a wrong assumption could change the result. Preserve the calculation and final cautious sentence for review.