Design Hydrology, Sediment and Site Investigations

Introduction and Context

Hydraulic structures are designed to perform safely under uncertain natural conditions. A spillway must pass floods without unacceptable damage; a reservoir must store water while maintaining required operating levels; an intake must remain functional despite sediment deposition; and a dam foundation must remain stable under seepage, loading and earthquake effects. These decisions depend on three interconnected investigations: hydrology, sediment assessment and site investigation.

Design hydrology converts rainfall, evaporation and streamflow records into design floods, hydrographs, reservoir levels and water-availability estimates. Sediment studies determine the quantity and character of material transported by a river and its likely effects on storage, outlets, turbines and channels. Site investigations establish the topographic, geological, geotechnical and environmental conditions governing the feasibility and safety of the structure.

Because observations are incomplete and natural systems are variable, design values are not exact predictions. They are risk-informed estimates based on data, models, assumptions and selected safety margins. Professional practice therefore requires explicit treatment of uncertainty rather than the presentation of a single number as unquestionable fact.

Learning Objectives

After completing this chapter, the reader should be able to:

1. distinguish design-flood concepts, return periods and risk;

2. select and apply appropriate frequency-analysis and rainfall–runoff methods;

3. construct and interpret design hydrographs and reservoir levels;

4. estimate sediment yield, sedimentation and trap efficiency;

5. plan topographic, geological, geotechnical and hydro-meteorological investigations;

6. evaluate uncertainty, data quality and model limitations; and

7. integrate investigation results into safe and economical hydraulic-structure decisions.

1. Design Hydrology for Hydraulic Structures

1.1 Hydrologic quantities and design criteria

The principal design quantities include:

• Design flood: a flood selected for a specified structure and consequence category.
• Peak discharge: the maximum instantaneous or mean discharge during a flood.
• Flood volume: the time integral of discharge above a selected base flow or threshold.
• Design hydrograph: discharge as a function of time, including peak, volume, duration and shape.
• Design storm: a rainfall event with specified depth, temporal pattern, spatial distribution and duration.
• Design water level: a reservoir or river level associated with a specified operating or flood condition.

The choice of design flood depends on failure consequences, hazard to downstream communities, dam classification, regulatory requirements, available data and the possibility of emergency action. Small diversion works may use a frequent design event, whereas a high-consequence dam may require a very rare event or a probable maximum flood (PMF), subject to national standards.

A return period, T, is the reciprocal of the annual exceedance probability, p:

T = 1/p

where p is expressed as a decimal and T is in years. A 100-year flood therefore has an annual exceedance probability of 0.01; it does not occur only once every 100 years. Assuming independent and stationary annual maxima, the probability that the event is exceeded at least once in n years is:

P = 1 - (1 - p)^n = 1 - (1 - 1/T)^n.

For T = 100 years and n = 50 years, P = 1 - 0.99^50 ≈ 0.395, or about 39.5%. This calculation is not valid without qualification where climate, catchment conditions or the definition of the event are nonstationary.

1.2 Data requirements and quality control

A defensible hydrologic study begins with a data inventory. It should include rainfall stations, stream gauges, evaporation records, reservoir levels, snow or glacier information where relevant, catchment characteristics, land-use history and records of floods not captured by gauges.

Quality control should address missing values, changes in gauge location or rating curves, instrument malfunction, inconsistent time bases, transcription errors and non-homogeneity. Streamflow is commonly derived from a stage–discharge rating curve. High-flow extrapolation beyond measured gaugings can produce substantial uncertainty and should be documented.

Records should be tested for consistency with nearby stations and for changes caused by reservoirs, diversions, urbanisation, deforestation or channel modification. Homogeneity tests are evidence, not proof; engineering judgment and historical knowledge remain necessary. Important extreme floods may be incorporated through historical information, but their dates, magnitudes and uncertainty must be recorded.

1.3 Flood-frequency analysis

For an annual-maximum series, one maximum discharge is selected for each water year. A probability distribution is then fitted to estimate quantiles such as Q100. Common distributions include the Gumbel extreme-value distribution, log-normal distribution and log-Pearson Type III distribution. The selected distribution should follow applicable national guidance and be checked using graphical diagnostics, goodness-of-fit measures, parameter uncertainty and physical plausibility.

The Gumbel model is often expressed as:

Q_T = μ + K_T σ,

where Q_T is the estimated discharge for return period T, μ is the sample mean, σ is the sample standard deviation and K_T is a frequency factor determined by the chosen Gumbel convention and T. The frequency factor must be calculated consistently with the adopted plotting-position and parameter definitions; tables from different conventions should not be mixed.

For a Gumbel variate, the reduced variate is:

y_T = -ln[-ln(1 - 1/T)].

With population parameters, the quantile is x_T = u + αy_T, where u is a location parameter and α is a scale parameter. In practice, sample estimators and finite-record corrections are used. The equation is appropriate only when the annual-maxima series is reasonably homogeneous and stationary and the model is adequate for the extrapolation range.

A peaks-over-threshold approach uses all independent exceedances above a threshold and may be more efficient than annual maxima, but it requires careful declustering and an appropriate exceedance distribution. Regional frequency analysis is valuable where a single gauge has a short record. It transfers information from hydrologically similar catchments using dimensionless growth curves or regional statistical models. Similarity must be demonstrated; geographical proximity alone is insufficient.

1.4 Rainfall–runoff transformation

Where streamflow records are inadequate, design floods may be developed from design storms. A common conceptual method is the Soil Conservation Service (SCS) curve-number method. The event runoff depth P_e is:

P_e = (P - I_a)^2/(P - I_a + S), for P > I_a;

P_e = 0, for P ≤ I_a,

where P is storm rainfall depth, I_a is initial abstraction and S is potential maximum retention, all in consistent depth units. In the standard SCS formulation, I_a = 0.2S and

S = 25400/CN - 254,

when S is in millimetres and CN is the dimensionless curve number. The method assumes a lumped catchment response and represents infiltration, interception and initial storage empirically. It is sensitive to antecedent moisture, land use, soil group, spatial rainfall and the selected CN; it should not be treated as a direct physical infiltration law.

The resulting runoff depth must be converted to a hydrograph. A unit hydrograph represents the direct-runoff hydrograph produced by one unit of effective rainfall uniformly distributed over the catchment for a specified duration. For a linear time-invariant catchment, convolution is used:

Q(t) = Σ P_e,i U(t - iΔt),

where P_e,i is effective rainfall during interval i and U is the corresponding unit-hydrograph ordinate. The assumptions of proportionality and time invariance become less reliable for very large storms, changing soil moisture, varying travel times and highly nonlinear catchments.

A rational estimate of peak runoff is:

Q_p = 0.278 C i A,

where Q_p is in m³/s, C is a dimensionless runoff coefficient, i is rainfall intensity in mm/h, and A is drainage area in km². The intensity should correspond approximately to the catchment time of concentration and the selected return period. The method is most appropriate for relatively small, homogeneous and urbanised catchments; it does not provide flood volume or a complete hydrograph without additional assumptions.

1.5 Design hydrographs and flood routing

A peak discharge alone is inadequate for spillway, outlet, floodplain and reservoir design. The hydrograph must represent peak, volume, duration and timing. It may be developed from observed analogue floods, a unit hydrograph, a rainfall–runoff model, regionalised hydrographs or a synthetic distribution constrained by catchment response.

Reservoir routing is based on continuity:

I - O = dS/dt,

where I is inflow, O is outflow and S is storage. For a time step Δt, a trapezoidal approximation is:

(S_2 - S_1)/Δt = (I_1 + I_2)/2 - (O_1 + O_2)/2.

The storage–elevation relationship and outlet/spillway discharge–elevation relationship are required. Routing results depend on the time step, initial reservoir level, gate operation, outlet geometry and the assumed inflow hydrograph. Emergency spillways should be assessed for blockage, erosion, tailwater effects and non-ideal operation.

1.6 Reservoir levels and water yield

Important levels include dead-storage level, minimum operating level, normal or full-reservoir level, surcharge level and maximum water level. Their definitions vary by project and regulation. The elevation–area–storage curve is obtained from topographic survey or a digital elevation model. It should be checked near the reservoir rim, where small vertical errors may imply large area or storage errors.

Water-supply analysis commonly uses a mass-balance equation:

S_{t+1} = S_t + I_t + P_tA_t - E_tA_t - R_t - D_t - L_t,

where S is storage, I is inflow, P is precipitation, E is evaporation, R is controlled release, D is demand or diversion, L is seepage and other losses, and A is the relevant water surface area. Units must be consistent, usually converting millimetres over area into cubic metres. Constraints include minimum and maximum storage and release-capacity limits.

Yield should be evaluated using historical simulation, stochastic sequences or both. Reliability, resilience and vulnerability are more informative than average yield alone. For example, reliability may be the fraction of periods in which demand is met; resilience describes the speed of recovery from failure; vulnerability measures the magnitude or duration of deficits. Climate nonstationarity requires scenario analysis rather than blind reliance on the historical sequence.

2. Sediment Processes and Design

2.1 Sediment sources and modes of transport

Sediment originates from hillslope erosion, channel-bank erosion, mass movements, construction, mining and glacial or volcanic processes. It is transported as bed load, suspended load and wash load. Bed load moves by rolling, sliding and saltation near the bed; suspended load is maintained by turbulence; wash load consists of fine material whose concentration is controlled primarily by catchment supply rather than bed exchange.

Sediment affects reservoir capacity, intake reliability, turbine abrasion, sluice operation, downstream channel stability, delta formation and environmental continuity. A sediment study should therefore consider not only mass but also grain-size distribution, mineralogy, density, shape, settling velocity, concentration variability and contaminants.

2.2 Estimating sediment load

For a measured concentration C_s and discharge Q, instantaneous suspended-sediment mass rate is:

M_s = C_s Q,

provided units are compatible. If C_s is in kg/m³ and Q in m³/s, M_s is kg/s. Daily load is M_s multiplied by 86,400 seconds per day. A rating relationship may be fitted as:

L = aQ^b,

where L is sediment load, Q is discharge and a and b are empirical coefficients. Logarithmic regression can be biased when transformed back to arithmetic units; a retransformation correction should be considered. Sampling must capture high-flow events, because a large proportion of annual sediment transport may occur during a few floods.

Bed-load samplers, acoustic instruments, turbidity sensors, suspended-sediment samplers, repeat bathymetric surveys and remote sensing can be combined. Each method has limitations: samplers disturb the flow or under-sample coarse particles, turbidity requires calibration, and bathymetry measures net deposition rather than all transported material.

2.3 Reservoir sedimentation and trap efficiency

The gross annual sediment volume may be approximated by:

V_s = M_s,annual/(ρ_b),

where M_s,annual is annual dry sediment mass and ρ_b is the deposited bulk dry density. Bulk density changes with grain size, consolidation, organic content and reservoir age; using a single value can cause significant error.

A reservoir’s trap efficiency is the fraction of incoming sediment retained. It depends on capacity, inflow, residence time, operation, density currents, particle size and outlet arrangements. Empirical methods such as Brune-type curves provide preliminary estimates for normally ponded reservoirs, but they are not universal laws. Detailed projects should use sediment routing, settling analysis, density-current assessment and repeated surveys.

Sediment management options include upstream watershed treatment, check dams, sediment bypass tunnels, sluicing, drawdown flushing, density-current venting, dredging, sediment excluders and improved intake placement. Each has hydraulic, environmental, operational and economic consequences. A sediment budget should distinguish incoming, deposited, released and remobilised material.

3. Site Investigations

3.1 Investigation philosophy

Site investigation is staged so that expensive and intrusive work is directed by progressively improved understanding. A typical sequence is desk study and remote sensing; reconnaissance; topographic and hydro-meteorological surveys; geological mapping and geophysics; exploratory drilling, trial pits and in-situ tests; laboratory testing; and construction-stage verification.

The investigation must address the structure, reservoir rim, abutments, borrow areas, spoil areas, access roads, diversion works, downstream channel and seismic sources. Investigation density should reflect geological variability and the consequences of uncertainty, not simply a fixed number of boreholes.

3.2 Topographic investigation

Topographic surveys establish contours, river cross-sections, longitudinal profiles, structure alignments, reservoir capacity and floodplain geometry. Total stations, GNSS, photogrammetry, LiDAR, multibeam bathymetry and satellite data may be integrated. Vertical and horizontal datums must be defined and preserved through permanent benchmarks.

A digital elevation model should report resolution, vertical accuracy, interpolation method, vegetation treatment and water-surface limitations. Survey errors can affect spillway crest elevation, freeboard, storage estimates and flood inundation maps. Independent checks and closure surveys are essential.

3.3 Geological investigation

Geological work identifies lithology, weathering, discontinuities, faults, joints, folds, karst, landslides, permeability contrasts and potential construction materials. Geological maps should be linked to engineering properties rather than treated as descriptive illustrations. Particular attention is required at abutments, spillway excavations, diversion tunnels, grout curtains and reservoir rims.

Geophysical methods such as seismic refraction, electrical resistivity, seismic tomography and ground-penetrating radar can identify contrasts and guide boreholes. They are indirect methods and require calibration against exposures and cores. Borehole logging should record recovery, rock quality designation, fracture condition, weathering and groundwater observations.

3.4 Geotechnical investigation and testing

Typical field tests include standard penetration testing, cone penetration testing, pressuremeter testing, permeability or packer testing, vane shear testing and plate-load testing. Laboratory tests may include grain-size analysis, Atterberg limits, compaction, triaxial or direct-shear strength, consolidation, permeability, intact-rock strength, point-load testing and durability tests.

Design parameters should distinguish total-stress and effective-stress conditions. For effective stress, shear strength is commonly represented as:

τ_f = c' + σ'_n tan φ',

where τ_f is shear strength at failure, c' is effective cohesion, σ'_n is effective normal stress and φ' is effective friction angle. The relationship is an idealised Mohr–Coulomb model; it is not valid without consideration of stress path, drainage, strain level, anisotropy, brittle behaviour and scale effects.

Groundwater investigation should measure piezometric levels through time, not only during drilling. Seepage gradients, uplift, leakage paths and seasonal variation strongly influence dam and foundation design. Pressure testing and tracer studies may be appropriate for fractured rock or karst.

3.5 Hydro-meteorological and environmental investigations

Rainfall, evaporation, temperature, wind, snow, sediment, floods and groundwater should be monitored at temporal resolutions suited to the design problem. A flood study may require sub-hourly rainfall and high-frequency stage records, whereas water-balance studies may use daily or monthly data.

Environmental investigation includes aquatic ecology, sediment quality, downstream geomorphic processes, wetlands, fisheries, cultural heritage and affected communities. Flow releases and sediment continuity may be regulatory requirements. Consultation, transparent disclosure and fair treatment of affected groups are part of technically responsible design, not optional public-relations activities.

4. Uncertainty, Risk and Decision Quality

Uncertainty has several forms. Aleatory variability is natural randomness, such as year-to-year rainfall variability. Epistemic uncertainty results from limited knowledge, such as uncertain rating curves or geological continuity. Model uncertainty arises from simplified rainfall–runoff, sediment and routing representations. Scenario uncertainty arises from future climate, land use and operations.

A useful workflow is to identify uncertain inputs, assign defensible ranges or probability distributions, propagate them through the model, test sensitivity and report results as ranges or probability-based metrics. Monte Carlo simulation can sample uncertain parameters, but its output is only as credible as the input distributions and model structure. Sensitivity analysis often reveals that a few factors—such as extreme rainfall, curve number, spillway blockage or sediment bulk density—dominate the decision.

Risk may be expressed conceptually as the combination of probability and consequence. Risk analysis should consider failure modes, not just return periods: overtopping, internal erosion, slope instability, foundation sliding, spillway erosion, blockage, landslide-generated waves and downstream inundation. Failure-mode analysis, independent checking, surveillance and emergency action planning should accompany numerical design.

5. Integrated Applied Case: Illustrative River-Valley Storage Project

Background

A proposed 42 m-high zoned earth dam is considered for municipal supply and flood attenuation in a 620 km² basin. The stream gauge has 34 years of daily data, but only 18 years include reliable high-flow measurements. The reservoir site lies downstream of highly weathered rock, and reconnaissance identifies active bank erosion and a small upstream landslide.

Problem

The design team must estimate the flood for spillway sizing, develop reservoir operating levels, assess storage loss from sediment and determine whether the foundation and abutments are suitable.

Analysis

Annual maxima were checked against nearby gauges and historical flood accounts. Rating-curve extrapolation was identified as a major uncertainty. A regional frequency analysis was compared with a fitted local distribution, and a rainfall–runoff model was calibrated against several observed events. Rather than selecting the largest unexamined value, the team developed a range of plausible hydrographs and routed each through the proposed reservoir.

Topographic LiDAR was checked by ground survey along the dam axis, spillway and reservoir rim. Drilling and seismic refraction indicated a shallow weathered zone over stronger rock, but one abutment contained a fractured, permeable zone. Packer tests and piezometers were installed to define seepage conditions.

Suspended-sediment sampling during five floods showed that sediment transport was strongly event-driven. A preliminary rating curve was used only for screening, and repeat bathymetric surveys were specified before final sediment-storage allocation. The upstream landslide was assessed as a potential episodic source rather than included as a constant annual yield.

Approach and solution

The selected design included a conservative spillway hydrograph envelope, gated flood routing with an independent emergency overflow path, increased monitoring of high flows and a dam-safety review of all assumptions. Sediment provisions included dead storage allocated from a range of bulk densities, an intake positioned above the expected delta front, and investigation of periodic sluicing. The fractured abutment required a grout curtain concept, drainage provisions and construction-stage verification.

Outcome and learning points

The project was not approved on the basis of one “best” hydrologic estimate. It was advanced because independent methods converged sufficiently for the key decisions, remaining uncertainty was linked to monitoring and operational actions, and failure consequences were explicitly considered. The case demonstrates that design hydrology, sediment assessment and site investigation must be integrated: a flood hydrograph affects sediment transport; sediment affects reservoir geometry; and geology controls seepage, construction and the reliability of hydraulic components.

6. Contemporary Tools and Best Practice

Cloud computing, geographic information systems, LiDAR, satellite rainfall, distributed hydrologic models, digital twins and machine-learning methods can improve spatial coverage and scenario analysis. Artificial intelligence may assist quality control, pattern recognition, gap filling and surrogate modelling. It must not replace physical reasoning, independent validation or professional accountability. Black-box predictions are unsafe when extrapolating beyond the training data, especially for rare floods.

Best practice includes maintaining a traceable data register; documenting assumptions and units; using independent methods; calibrating models against multiple events; separating calibration from validation; conducting sensitivity and uncertainty analysis; checking extreme-value extrapolation; surveying sediment repeatedly; integrating geology with design; and using independent peer review for high-consequence structures. Monitoring plans should be designed during investigation, with trigger levels and response actions defined before operation.

Chapter Summary

Design hydrology transforms uncertain rainfall and flow information into design floods, hydrographs, reservoir levels and water-yield estimates. Return period is a probability concept, not a fixed recurrence interval, and multi-year risk must be calculated with explicit assumptions. Frequency analysis, rainfall–runoff modelling, unit hydrographs and reservoir routing each have defined data requirements and limitations.

Sediment assessment requires understanding sources, transport modes, concentration–discharge behaviour, grain size, bulk density and reservoir trap efficiency. Sedimentation affects capacity, outlets, turbines, environmental flows and long-term project performance. Site investigations proceed from desk study through reconnaissance, surveying, geophysics, drilling, testing and construction verification. Topography, geology, groundwater, geotechnical properties and hydro-meteorology must be interpreted together.

Uncertainty is unavoidable but can be managed through data-quality control, sensitivity analysis, probabilistic methods, scenario assessment, independent review, monitoring and adaptive operation. Safe hydraulic-structure design is therefore an evidence-based decision process rather than the calculation of a single supposedly exact value.

Key Terms

Design flood: Flood selected for the design or safety assessment of a structure.

Annual exceedance probability: Probability that a specified event is exceeded in one year.

Return period: Reciprocal of annual exceedance probability under the stated assumptions.

Design hydrograph: Discharge-time representation of a design flood.

Flood routing: Calculation of changing discharge and storage as a flood moves through a system.

Unit hydrograph: Direct-runoff hydrograph caused by one unit of effective rainfall of specified duration.

Curve number: Empirical parameter representing runoff response in the SCS method.

Water yield: Usable supply obtained from a reservoir under specified reliability criteria.

Sediment load: Mass rate or mass quantity of sediment transported by flowing water.

Bed load: Sediment moving close to the channel bed.

Suspended load: Sediment maintained in the water column by turbulence.

Trap efficiency: Fraction of incoming sediment retained in a reservoir.

Bulk density: Dry sediment mass per unit volume of deposited sediment.

Piezometric level: Hydraulic head measured in a piezometer or groundwater system.

Rock quality designation: Core-based index describing the proportion of sound rock pieces of specified length.

Effective stress: Stress carried by the soil skeleton and governing many strength and deformation behaviours.

Epistemic uncertainty: Uncertainty caused by incomplete knowledge or limited data.

Aleatory variability: Natural variability inherent in the process being studied.

Failure-mode analysis: Structured assessment of possible mechanisms leading to loss of function or safety.

Review Questions

Short Answer Questions

1. Why is a 100-year flood not expected to occur exactly once every 100 years?

2. State the continuity equation used in reservoir routing and define its terms.

3. Distinguish bed load from suspended load.

4. What information is obtained from a reservoir elevation–area–storage curve?

5. Why should groundwater levels be monitored over time during site investigation?

Long Answer Questions

1. Explain the complete procedure for developing a design flood from streamflow data, including data screening, frequency analysis and uncertainty assessment.

2. Discuss the assumptions, applications and limitations of the SCS curve-number method and unit-hydrograph approach.

3. Describe how sediment is measured and converted into an estimate of reservoir storage loss.

4. Explain the staged investigation process for a dam site and show how geological and geotechnical findings influence design.

5. Discuss how nonstationarity, climate change and model uncertainty should be incorporated into reservoir and spillway decisions.

Analytical / Application Questions

1. A 100-year event has annual exceedance probability 0.01. Assuming independence and stationarity, calculate the probability of at least one exceedance in 50 years and explain the engineering meaning of the result.

2. A catchment receives 120 mm of rainfall. Using the SCS method with CN = 75 and I_a = 0.2S, calculate the event runoff depth. State the assumptions and explain why the result should not be interpreted as a directly measured runoff value.

3. A sediment concentration of 0.8 kg/m³ occurs at a discharge of 250 m³/s for six hours. Calculate the transported sediment mass during that period and identify two reasons why extrapolating this observation to annual sediment yield may be unreliable.

Multiple Choice Questions with Answers

Question 1

A flood with a 2% annual exceedance probability has a return period of:

A. 2 years

B. 20 years

C. 50 years

D. 100 years

Correct Answer: C. 50 years

Question 2

For independent annual events, the probability of at least one exceedance in n years is:

A. (1-p)^n

B. 1-(1-p)^n

C. p/n

D. 1-pn

Correct Answer: B. 1-(1-p)^n

Question 3

In reservoir routing, the continuity relationship is:

A. I+O=dS/dt

B. I-O=dS/dt

C. O-I=dQ/dt

D. I/O=dS/dt

Correct Answer: B. I-O=dS/dt

Question 4

In the SCS curve-number method, S = 25400/CN - 254 gives S in:

A. metres

B. cubic metres

C. millimetres

D. millimetres per hour

Correct Answer: C. millimetres

Question 5

The rational-method peak discharge equation Qp = 0.278CiA uses rainfall intensity i in:

A. mm/h

B. m/s

C. m/day

D. mm/day

Correct Answer: A. mm/h

Question 6

Which sediment mode moves primarily near the channel bed by rolling, sliding or saltation?

A. Wash load

B. Bed load

C. Dissolved load

D. Atmospheric load

Correct Answer: B. Bed load

Question 7

If sediment concentration is 0.5 kg/m³ and discharge is 100 m³/s, the instantaneous sediment mass rate is:

A. 5 kg/s

B. 50 kg/s

C. 500 kg/s

D. 5000 kg/s

Correct Answer: B. 50 kg/s

Question 8

Which investigation most directly measures hydraulic head in a foundation?

A. Piezometer installation

B. Grain-size analysis

C. LiDAR survey

D. Point-load testing

Correct Answer: A. Piezometer installation

Question 9

A major limitation of a black-box AI model for rare-flood prediction is that it may:

A. Always conserve mass

B. Extrapolate reliably beyond its training data

C. Perform poorly outside the range represented by its training data

D. Eliminate model uncertainty

Correct Answer: C. Perform poorly outside the range represented by its training data

Question 10

Effective-stress Mohr–Coulomb strength is commonly expressed as:

A. τf = c' + σ'n tan φ'

B. τf = c'σ'n + tan φ'

C. τf = σ'n/c' + φ'

D. τf = c' - σ'n tan φ'

Correct Answer: A. τf = c' + σ'n tan φ'