Hydraulic Modeling, Dimensional Analysis and Similitude

This unit addresses a linked set of civil-engineering decisions rather than a collection of disconnected equations. In a scaled bridge-abutment experiment, an analyst first asks which processes dominate, where the control boundaries belong, and which field evidence can discriminate between alternative explanations. The governing relation is useful only after each variable, units, sign convention and physical domain have been checked. A calculation is one element of a design argument, never a substitute for the argument itself.

Conservation principles and engineering judgment

Mass, momentum and mechanical energy are the principal organizing principles of this unit. Their apparent simplicity hides the hardest practical question: which losses, forces, storage terms and boundary conditions belong inside the model? In a prototype urban-drainage model, a visually convincing plot may still be wrong if an outlet is submerged when the model assumes free flow. Start by drawing the geometry, marking the upstream and downstream reference elevations, identifying the expected flow regime and recording the source of every coefficient.

Dimensional checks expose many errors before they become design mistakes. Keep discharge in cubic meters per second when deriving SI power or forces; distinguish kPa from Pa and diameter from radius; label gauge and absolute pressure; and state whether the model uses a Darcy or a Fanning friction factor. Any reported precision should reflect input uncertainty, not the number of digits available in a computer.

An engineering investigation also has an ethical and operational dimension. If a flume calibration campaign carries contaminated water or serves a public-facing facility, failure is not only an algebraic concern: unexpected overtopping, pressure loss, scour or cavitation can harm people and ecosystems. Use the small models here to discover sensitivities and build intuition. Actual selection, construction or operation requires qualified review, site data and applicable standards.

Lesson 8.1 — Dimensional analysis and similitude

Physical basis. Dimensionless groups reveal which physical phenomena dominate a laboratory model. Reynolds similarity is relevant to viscous internal flow; Froude similarity governs gravity-dominated free-surface motion. A small model often cannot match both simultaneously.

Representative relation: Fr=V/sqrt(gL); Re=VL/nu.

Assumptions and failure modes. A physical model with correct geometry may still have misleading surface-tension or viscous effects. Document the similitude that was prioritized and the distortions accepted.

Investigation objective. Derive the Froude scaling of velocity, discharge and time when the geometric length ratio is specified.

Why the model matters for an actual project

The decision in a scaled bridge-abutment experiment depends on more than a single value of an unknown. Begin with an explicit sketch and a table listing the fluid state, channel or pipe geometry, the selected datum, boundary elevations and the intended output. Record the difference between a measured input and one chosen for an exploratory calculation. This distinction becomes essential when an optimistic coefficient produces a result that appears to satisfy a design target.

For dimensional analysis and similitude, the equation Fr=V/sqrt(gL); Re=VL/nu can be used as a starting point, not a universal law independent of context. First trace every term back to the conservation principle or constitutive relation that supplies it. Next imagine reversing the direction of flow, doubling the characteristic dimension or moving the gauge datum. Some predictions must change in a specific direction; if the spreadsheet or simulation does not respect that monotonicity, stop and inspect the setup.

How to verify the result

For a 1D Saint-Venant solver trial, collect a downstream tailwater survey. A good verification uses at least one independently observable quantity, not the same calculation rearranged. Where a coefficient is empirical, report its source and the regime in which it was obtained. If the measured result differs from the model, examine datum errors, flow regime and equipment condition before altering a parameter merely to improve the fit.

Interpretation and limitations

A physical model with correct geometry may still have misleading surface-tension or viscous effects. Document the similitude that was prioritized and the distortions accepted. In addition, operating schedule may vary with operating season or maintenance. A calculated response should therefore be accompanied by a bounded sensitivity study: change one influential variable, keep the others fixed, document whether the output rises or falls, and identify a physical observation that could falsify the assumed model. Engineers who cannot explain the direction of that change are not ready to use the output for selection.

Stepwise numerical practice — Froude velocity scaling

The repeatable computation below uses prototype/model velocity ratio = sqrt(length ratio) and the relationship Vr=sqrt(Lr). Each row has a different operating situation. Students should first reproduce the central case by hand, then use code or a calculator to vary one input and check dimensional consistency. Because the data are synthetic, they are not evidence that any real plant or channel satisfies performance requirements.

Case 8.1.1: A prototype urban-drainage model. Inputs: Geometric scale ratio = 10 –. The predicted 3.1623 dimensionless comes from substituting these measurements into Vr=sqrt(Lr) with consistent SI conversions. Because boundary-condition uncertainty can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a controlled flow-stepping trial as the external validation measurement.

Case 8.1.2: A flume calibration campaign. Inputs: Geometric scale ratio = 12 –. Using Vr=sqrt(Lr), the worked output is 3.4641 dimensionless. A plausible error in operating schedule may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan an independently calibrated pressure logger rather than calibrating the equation only against its own output.

Case 8.1.3: A 1d saint-venant solver trial. Inputs: Geometric scale ratio = 14 –. The governing computation gives 3.7417 dimensionless. Sketch the geometry and trace how each input enters Vr=sqrt(Lr). To avoid false precision, perturb the least certain input, discuss how upstream water level affects the chosen boundary condition, and recommend a seasonal water-level record to challenge the prediction independently.

Case 8.1.4: A rural irrigation sensitivity study. Inputs: Geometric scale ratio = 16 –. The predicted 4 dimensionless comes from substituting these measurements into Vr=sqrt(Lr) with consistent SI conversions. Because surface roughness can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a repeatable instrument zero check as the external validation measurement.

Case 8.1.5: A coastal gate physical model. Inputs: Geometric scale ratio = 18 –. Using Vr=sqrt(Lr), the worked output is 4.2426 dimensionless. A plausible error in sediment accumulation may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan a controlled flow-stepping trial rather than calibrating the equation only against its own output.

Case 8.1.6: An emergency overtopping scenario. Inputs: Geometric scale ratio = 20 –. The governing computation gives 4.4721 dimensionless. Sketch the geometry and trace how each input enters Vr=sqrt(Lr). To avoid false precision, perturb the least certain input, discuss how maintenance accessibility affects the chosen boundary condition, and recommend an independently calibrated pressure logger to challenge the prediction independently.

Case 8.1.7: A field-model reconciliation workshop. Inputs: Geometric scale ratio = 22 –. The predicted 4.6904 dimensionless comes from substituting these measurements into Vr=sqrt(Lr) with consistent SI conversions. Because seasonal temperature change can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a seasonal water-level record as the external validation measurement.

Concept and transfer questions

1. Starting with a control-volume or free-body diagram, explain why Fr=V/sqrt(gL); Re=VL/nu follows from the stated assumptions.
2. In a coastal gate physical model, identify two reasons a measured value could differ from this simplified estimate.
3. Describe a feasible calibration or field test that would bound uncertainty in upstream water level.
4. Give an example in which use of the relationship outside its validity range could reverse a project decision.

Mini-design exercise. For an emergency overtopping scenario, specify the physical layout, relevant upstream and downstream boundary data, the likely controlling mechanism, a conservative operating scenario, and the evidence that would be required before approving a change. Give an uncertainty range, not only a single final number.

Lesson 8.2 — Physical models and instrument plans

Physical basis. A physical model is an experiment with scaling hypotheses. Constructing a reliable model requires a plan for boundary control, calibration, repeatability, recording of water levels, and reconciliation of forces and discharge.

Representative relation: V_r=sqrt(L_r); Q_r=L_r^(5/2) (Froude, geometric similarity).

Assumptions and failure modes. A mismatch between model and prototype roughness can change the observed water surface even when Froude number is correctly matched.

Investigation objective. Specify measurement stations and an acceptance criterion for a reduced-scale bridge-waterway model.

Why the model matters for an actual project

The decision in a prototype urban-drainage model depends on more than a single value of an unknown. Begin with an explicit sketch and a table listing the fluid state, channel or pipe geometry, the selected datum, boundary elevations and the intended output. Record the difference between a measured input and one chosen for an exploratory calculation. This distinction becomes essential when an optimistic coefficient produces a result that appears to satisfy a design target.

For physical models and instrument plans, the equation V_r=sqrt(L_r); Q_r=L_r^(5/2) (Froude, geometric similarity) can be used as a starting point, not a universal law independent of context. First trace every term back to the conservation principle or constitutive relation that supplies it. Next imagine reversing the direction of flow, doubling the characteristic dimension or moving the gauge datum. Some predictions must change in a specific direction; if the spreadsheet or simulation does not respect that monotonicity, stop and inspect the setup.

How to verify the result

For a rural irrigation sensitivity study, collect a controlled flow-stepping trial. A good verification uses at least one independently observable quantity, not the same calculation rearranged. Where a coefficient is empirical, report its source and the regime in which it was obtained. If the measured result differs from the model, examine datum errors, flow regime and equipment condition before altering a parameter merely to improve the fit.

Interpretation and limitations

A mismatch between model and prototype roughness can change the observed water surface even when Froude number is correctly matched. In addition, upstream water level may vary with operating season or maintenance. A calculated response should therefore be accompanied by a bounded sensitivity study: change one influential variable, keep the others fixed, document whether the output rises or falls, and identify a physical observation that could falsify the assumed model. Engineers who cannot explain the direction of that change are not ready to use the output for selection.

Stepwise numerical practice — Froude discharge scaling

The repeatable computation below uses discharge ratio = geometric length ratio to 5/2 and the relationship Qr=Lr^2.5. Each row has a different operating situation. Students should first reproduce the central case by hand, then use code or a calculator to vary one input and check dimensional consistency. Because the data are synthetic, they are not evidence that any real plant or channel satisfies performance requirements.

Case 8.2.1: A flume calibration campaign. Inputs: Geometric scale ratio = 9 –. The predicted 243 dimensionless comes from substituting these measurements into Qr=Lr^2.5 with consistent SI conversions. Because operating schedule can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a survey of cross-section geometry as the external validation measurement.

Case 8.2.2: A 1d saint-venant solver trial. Inputs: Geometric scale ratio = 11 –. Using Qr=Lr^2.5, the worked output is 401.31 dimensionless. A plausible error in upstream water level may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan a timed volumetric collection measurement rather than calibrating the equation only against its own output.

Case 8.2.3: A rural irrigation sensitivity study. Inputs: Geometric scale ratio = 13 –. The governing computation gives 609.34 dimensionless. Sketch the geometry and trace how each input enters Qr=Lr^2.5. To avoid false precision, perturb the least certain input, discuss how surface roughness affects the chosen boundary condition, and recommend a field roughness estimate to challenge the prediction independently.

Case 8.2.4: A coastal gate physical model. Inputs: Geometric scale ratio = 15 –. The predicted 871.42 dimensionless comes from substituting these measurements into Qr=Lr^2.5 with consistent SI conversions. Because sediment accumulation can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a downstream tailwater survey as the external validation measurement.

Case 8.2.5: An emergency overtopping scenario. Inputs: Geometric scale ratio = 17 –. Using Qr=Lr^2.5, the worked output is 1191.6 dimensionless. A plausible error in maintenance accessibility may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan a survey of cross-section geometry rather than calibrating the equation only against its own output.

Case 8.2.6: A field-model reconciliation workshop. Inputs: Geometric scale ratio = 19 –. The governing computation gives 1573.6 dimensionless. Sketch the geometry and trace how each input enters Qr=Lr^2.5. To avoid false precision, perturb the least certain input, discuss how seasonal temperature change affects the chosen boundary condition, and recommend a timed volumetric collection measurement to challenge the prediction independently.

Case 8.2.7: A scaled bridge-abutment experiment. Inputs: Geometric scale ratio = 21 –. The predicted 2020.9 dimensionless comes from substituting these measurements into Qr=Lr^2.5 with consistent SI conversions. Because instrument bias can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a field roughness estimate as the external validation measurement.

Concept and transfer questions

1. Starting with a control-volume or free-body diagram, explain why V_r=sqrt(L_r); Q_r=L_r^(5/2) (Froude, geometric similarity) follows from the stated assumptions.
2. In an emergency overtopping scenario, identify two reasons a measured value could differ from this simplified estimate.
3. Describe a feasible calibration or field test that would bound uncertainty in surface roughness.
4. Give an example in which use of the relationship outside its validity range could reverse a project decision.

Mini-design exercise. For a field-model reconciliation workshop, specify the physical layout, relevant upstream and downstream boundary data, the likely controlling mechanism, a conservative operating scenario, and the evidence that would be required before approving a change. Give an uncertainty range, not only a single final number.

Lesson 8.3 — Numerical discretization and conservation

Physical basis. Numerical solutions approximate conservation equations on discrete cells or nodes. Finite-volume methods emphasize flux balance; a stable time step may be constrained by flow speed and gravity-wave celerity. Grid convergence is a separate question from solver convergence.

Representative relation: CFL=(|V|+sqrt(gy))Δt/Δx (illustrative 1D).

Assumptions and failure modes. A solution that converges algebraically may still be wrong because of coarse geometry, incorrect boundaries or the wrong turbulence/roughness assumptions.

Investigation objective. Compare coarse and refined cell sizes, calculate mass-balance residuals and document numerical uncertainty.

Why the model matters for an actual project

The decision in a flume calibration campaign depends on more than a single value of an unknown. Begin with an explicit sketch and a table listing the fluid state, channel or pipe geometry, the selected datum, boundary elevations and the intended output. Record the difference between a measured input and one chosen for an exploratory calculation. This distinction becomes essential when an optimistic coefficient produces a result that appears to satisfy a design target.

For numerical discretization and conservation, the equation CFL=(|V|+sqrt(gy))Δt/Δx (illustrative 1D) can be used as a starting point, not a universal law independent of context. First trace every term back to the conservation principle or constitutive relation that supplies it. Next imagine reversing the direction of flow, doubling the characteristic dimension or moving the gauge datum. Some predictions must change in a specific direction; if the spreadsheet or simulation does not respect that monotonicity, stop and inspect the setup.

How to verify the result

For a coastal gate physical model, collect a survey of cross-section geometry. A good verification uses at least one independently observable quantity, not the same calculation rearranged. Where a coefficient is empirical, report its source and the regime in which it was obtained. If the measured result differs from the model, examine datum errors, flow regime and equipment condition before altering a parameter merely to improve the fit.

Interpretation and limitations

A solution that converges algebraically may still be wrong because of coarse geometry, incorrect boundaries or the wrong turbulence/roughness assumptions. In addition, surface roughness may vary with operating season or maintenance. A calculated response should therefore be accompanied by a bounded sensitivity study: change one influential variable, keep the others fixed, document whether the output rises or falls, and identify a physical observation that could falsify the assumed model. Engineers who cannot explain the direction of that change are not ready to use the output for selection.

Stepwise numerical practice — time-step Courant screen

The repeatable computation below uses cfl = (v+sqrt(gy))dt/dx and the relationship CFL=(V+sqrt(gy))dt/dx. Each row has a different operating situation. Students should first reproduce the central case by hand, then use code or a calculator to vary one input and check dimensional consistency. Because the data are synthetic, they are not evidence that any real plant or channel satisfies performance requirements.

Case 8.3.1: A 1d saint-venant solver trial. Inputs: Velocity = 0.95 m/s; Depth = 1.35 m; Time step = 0.1 s; Cell length = 1.5 m. The predicted 0.30594 dimensionless comes from substituting these measurements into CFL=(V+sqrt(gy))dt/dx with consistent SI conversions. Because upstream water level can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain an independently calibrated pressure logger as the external validation measurement.

Case 8.3.2: A rural irrigation sensitivity study. Inputs: Velocity = 1 m/s; Depth = 1.4 m; Time step = 0.1 s; Cell length = 1.5 m. Using CFL=(V+sqrt(gy))dt/dx, the worked output is 0.31373 dimensionless. A plausible error in surface roughness may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan a seasonal water-level record rather than calibrating the equation only against its own output.

Case 8.3.3: A coastal gate physical model. Inputs: Velocity = 1.05 m/s; Depth = 1.45 m; Time step = 0.1 s; Cell length = 1.5 m. The governing computation gives 0.32144 dimensionless. Sketch the geometry and trace how each input enters CFL=(V+sqrt(gy))dt/dx. To avoid false precision, perturb the least certain input, discuss how sediment accumulation affects the chosen boundary condition, and recommend a repeatable instrument zero check to challenge the prediction independently.

Case 8.3.4: An emergency overtopping scenario. Inputs: Velocity = 1.1 m/s; Depth = 1.5 m; Time step = 0.1 s; Cell length = 1.5 m. The predicted 0.32907 dimensionless comes from substituting these measurements into CFL=(V+sqrt(gy))dt/dx with consistent SI conversions. Because maintenance accessibility can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a controlled flow-stepping trial as the external validation measurement.

Case 8.3.5: A field-model reconciliation workshop. Inputs: Velocity = 1.15 m/s; Depth = 1.55 m; Time step = 0.1 s; Cell length = 1.5 m. Using CFL=(V+sqrt(gy))dt/dx, the worked output is 0.33663 dimensionless. A plausible error in seasonal temperature change may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan an independently calibrated pressure logger rather than calibrating the equation only against its own output.

Case 8.3.6: A scaled bridge-abutment experiment. Inputs: Velocity = 1.2 m/s; Depth = 1.6 m; Time step = 0.1 s; Cell length = 1.5 m. The governing computation gives 0.34412 dimensionless. Sketch the geometry and trace how each input enters CFL=(V+sqrt(gy))dt/dx. To avoid false precision, perturb the least certain input, discuss how instrument bias affects the chosen boundary condition, and recommend a seasonal water-level record to challenge the prediction independently.

Case 8.3.7: A prototype urban-drainage model. Inputs: Velocity = 1.25 m/s; Depth = 1.65 m; Time step = 0.1 s; Cell length = 1.5 m. The predicted 0.35155 dimensionless comes from substituting these measurements into CFL=(V+sqrt(gy))dt/dx with consistent SI conversions. Because boundary-condition uncertainty can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a repeatable instrument zero check as the external validation measurement.

Concept and transfer questions

1. Starting with a control-volume or free-body diagram, explain why CFL=(|V|+sqrt(gy))Δt/Δx (illustrative 1D) follows from the stated assumptions.
2. In a field-model reconciliation workshop, identify two reasons a measured value could differ from this simplified estimate.
3. Describe a feasible calibration or field test that would bound uncertainty in sediment accumulation.
4. Give an example in which use of the relationship outside its validity range could reverse a project decision.

Mini-design exercise. For a scaled bridge-abutment experiment, specify the physical layout, relevant upstream and downstream boundary data, the likely controlling mechanism, a conservative operating scenario, and the evidence that would be required before approving a change. Give an uncertainty range, not only a single final number.

Lesson 8.4 — Uncertainty, calibration and decisions

Physical basis. Calibration adjusts model parameters to observations; validation evaluates performance on independent information where possible. Sensitivity analysis varies uncertain inputs; scenario analysis evaluates distinct future operating conditions. Neither can prove an unsafe design safe.

Representative relation: y=f(x,θ); residual=observed-predicted.

Assumptions and failure modes. Avoid calibrating multiple parameters to one observation and claiming uniqueness. An apparently good fit can hide compensating errors.

Investigation objective. Prepare an uncertainty register with parameter ranges, evidence sources, sensitivity ranking and management responses.

Why the model matters for an actual project

The decision in a 1D Saint-Venant solver trial depends on more than a single value of an unknown. Begin with an explicit sketch and a table listing the fluid state, channel or pipe geometry, the selected datum, boundary elevations and the intended output. Record the difference between a measured input and one chosen for an exploratory calculation. This distinction becomes essential when an optimistic coefficient produces a result that appears to satisfy a design target.

For uncertainty, calibration and decisions, the equation y=f(x,θ); residual=observed-predicted can be used as a starting point, not a universal law independent of context. First trace every term back to the conservation principle or constitutive relation that supplies it. Next imagine reversing the direction of flow, doubling the characteristic dimension or moving the gauge datum. Some predictions must change in a specific direction; if the spreadsheet or simulation does not respect that monotonicity, stop and inspect the setup.

How to verify the result

For an emergency overtopping scenario, collect an independently calibrated pressure logger. A good verification uses at least one independently observable quantity, not the same calculation rearranged. Where a coefficient is empirical, report its source and the regime in which it was obtained. If the measured result differs from the model, examine datum errors, flow regime and equipment condition before altering a parameter merely to improve the fit.

Interpretation and limitations

Avoid calibrating multiple parameters to one observation and claiming uniqueness. An apparently good fit can hide compensating errors. In addition, sediment accumulation may vary with operating season or maintenance. A calculated response should therefore be accompanied by a bounded sensitivity study: change one influential variable, keep the others fixed, document whether the output rises or falls, and identify a physical observation that could falsify the assumed model. Engineers who cannot explain the direction of that change are not ready to use the output for selection.

Stepwise numerical practice — simple residual calculation

The repeatable computation below uses residual = observation minus prediction and the relationship r=y_obs−y_calc. Each row has a different operating situation. Students should first reproduce the central case by hand, then use code or a calculator to vary one input and check dimensional consistency. Because the data are synthetic, they are not evidence that any real plant or channel satisfies performance requirements.

Case 8.4.1: A rural irrigation sensitivity study. Inputs: Observed level = 4.3 m; Predicted level = 3.96 m. The predicted 0.34 m comes from substituting these measurements into r=y_obs−y_calc with consistent SI conversions. Because surface roughness can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a timed volumetric collection measurement as the external validation measurement.

Case 8.4.2: A coastal gate physical model. Inputs: Observed level = 4.5 m; Predicted level = 4.15 m. Using r=y_obs−y_calc, the worked output is 0.35 m. A plausible error in sediment accumulation may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan a field roughness estimate rather than calibrating the equation only against its own output.

Case 8.4.3: An emergency overtopping scenario. Inputs: Observed level = 4.7 m; Predicted level = 4.34 m. The governing computation gives 0.36 m. Sketch the geometry and trace how each input enters r=y_obs−y_calc. To avoid false precision, perturb the least certain input, discuss how maintenance accessibility affects the chosen boundary condition, and recommend a downstream tailwater survey to challenge the prediction independently.

Case 8.4.4: A field-model reconciliation workshop. Inputs: Observed level = 4.9 m; Predicted level = 4.53 m. The predicted 0.37 m comes from substituting these measurements into r=y_obs−y_calc with consistent SI conversions. Because seasonal temperature change can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a survey of cross-section geometry as the external validation measurement.

Case 8.4.5: A scaled bridge-abutment experiment. Inputs: Observed level = 5.1 m; Predicted level = 4.72 m. Using r=y_obs−y_calc, the worked output is 0.38 m. A plausible error in instrument bias may dominate the last displayed digit. Show each intermediate step, then state whether a larger first listed parameter would move the result in the expected physical direction. Plan a timed volumetric collection measurement rather than calibrating the equation only against its own output.

Case 8.4.6: A prototype urban-drainage model. Inputs: Observed level = 5.3 m; Predicted level = 4.91 m. The governing computation gives 0.39 m. Sketch the geometry and trace how each input enters r=y_obs−y_calc. To avoid false precision, perturb the least certain input, discuss how boundary-condition uncertainty affects the chosen boundary condition, and recommend a field roughness estimate to challenge the prediction independently.

Case 8.4.7: A flume calibration campaign. Inputs: Observed level = 5.5 m; Predicted level = 5.1 m. The predicted 0.4 m comes from substituting these measurements into r=y_obs−y_calc with consistent SI conversions. Because operating schedule can shift in service, evaluate a second condition before interpreting the point estimate as a robust capacity. Obtain a downstream tailwater survey as the external validation measurement.

Concept and transfer questions

1. Starting with a control-volume or free-body diagram, explain why y=f(x,θ); residual=observed-predicted follows from the stated assumptions.
2. In a scaled bridge-abutment experiment, identify two reasons a measured value could differ from this simplified estimate.
3. Describe a feasible calibration or field test that would bound uncertainty in maintenance accessibility.
4. Give an example in which use of the relationship outside its validity range could reverse a project decision.

Mini-design exercise. For a prototype urban-drainage model, specify the physical layout, relevant upstream and downstream boundary data, the likely controlling mechanism, a conservative operating scenario, and the evidence that would be required before approving a change. Give an uncertainty range, not only a single final number.

Unit 8 applied investigation

Your group has been assigned a field-model reconciliation workshop. Develop a falsifiable hydraulic model, list all inputs in a measurement register, show how mass conservation is checked and identify a second method of checking the primary output. Compare a low-flow, baseline and high-flow condition. Discuss operating schedule and sediment accumulation as two possible sources of disagreement. Provide a table of calculated values, explain whether results are stable under small input changes, and separate teaching-model conclusions from claims requiring field verification.

Decision memo

Prepare a two-page report containing the question, assumptions, data provenance, dimensional checks, one independent cross-check, sensitivity findings, limitations and a practical next measurement. Report uncertainty without inventing probabilities. Include a safe operating implication and identify exactly which engineering decision is not justified by the simplified model.

Review prompts

Explain why conservation is not equivalent to assuming zero losses. Distinguish an ideal law, an empirical coefficient, an instrument reading and a numerical approximation. Explain how a model can generate a precise value yet still be inappropriate for a culvert, pump, gate or open-channel condition.