Hydrological Data, Analysis and Uncertainty

Introduction and Context

Water-resources decisions depend on observations of a highly variable natural system. Rainfall, evaporation, infiltration, soil moisture, streamflow and groundwater levels fluctuate across time and space, while measurements are affected by instrument limitations, sampling choices and changing environmental conditions. Hydrological analysis therefore involves two linked tasks: estimating the behaviour of the water cycle from data and expressing the uncertainty associated with those estimates.

This chapter develops the principal methods used to analyse hydrological observations for planning, design and management. It considers data acquisition and quality control, precipitation analysis, evaporation and infiltration estimation, rainfall–runoff relationships, streamflow and hydrograph analysis, probability distributions, frequency analysis and return periods. Particular attention is given to assumptions, data limitations and the distinction between natural variability and measurement or model uncertainty.

Hydrological results are used in reservoir operation, flood protection, drainage design, irrigation planning, drought management, environmental-flow assessment and climate adaptation. A technically sound analysis must therefore be transparent, reproducible and proportionate to the consequences of error.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Describe the principal types and sources of hydrological data.

2. Apply basic procedures for data screening, consistency checking and gap treatment.

3. Analyse precipitation, evaporation, infiltration, runoff and streamflow observations.

4. Interpret hydrograph components and extract useful indices.

5. Select and apply probability distributions and frequency-analysis methods.

6. Distinguish exceedance probability, non-exceedance probability and return period.

7. Quantify and communicate uncertainty in hydrological estimates.

8. Evaluate the suitability and limitations of hydrological analyses for engineering decisions.

1. Hydrological Data and the Measurement System

1.1 Hydrological variables

Common variables include:

• precipitation depth or intensity;
• air temperature, humidity, wind speed and solar radiation;
• evaporation or evapotranspiration;
• soil moisture and infiltration rate;
• river stage and discharge;
• groundwater level;
• sediment concentration and water quality indicators.

A hydrological record has at least four attributes: location, time step, measurement method and measurement uncertainty. A daily rainfall value, for example, is not interchangeable with an hourly intensity, even if both are expressed in millimetres.

1.2 Monitoring networks and instruments

Rainfall is measured using storage gauges, weighing gauges or tipping-bucket gauges. Gauge exposure, wind effects, splash, evaporation and blockage can create systematic errors. Radar and satellite products provide spatial information but require local calibration and may contain errors caused by beam blockage, bright-band effects, terrain and retrieval algorithms.

Streamflow is usually obtained by measuring stage and converting it to discharge through a stage–discharge relationship, or rating curve. The general rating relation is often represented as

Q = a(h - h_0)^b

where Q is discharge, h is measured stage, h0 is the stage corresponding approximately to zero flow, and a and b are fitted parameters. This relation is not universal. It may change when the channel geometry, roughness, control section or backwater condition changes. High flows are frequently extrapolated beyond the range of direct measurements and consequently have greater uncertainty.

Evaporation may be measured using pans, lysimeters or energy-balance instruments. Potential evapotranspiration is commonly estimated from meteorological data using physically based or empirical methods. Infiltration can be measured with infiltrometers, rainfall simulators or inferred from catchment water balances. Each method represents a different scale and set of boundary conditions.

1.3 Data quality control

Quality assurance begins before field collection. Stations should have documented coordinates, datum, instrument type, maintenance history, observation times and changes in land use or channel condition. Quality control commonly includes:

• range checks for physically implausible values;
• checks for duplicated timestamps and missing records;
• comparison with neighbouring stations;
• inspection of abrupt shifts or suspiciously repeated values;
• review of rating-curve changes;
• comparison of rainfall totals with storm reports and radar products;
• preservation of raw data separately from corrected data.

A flagged value should not automatically be deleted. It should be classified as valid, corrected, estimated or missing, with the reason recorded. Data provenance is essential for auditability.

1.4 Homogeneity, consistency and gaps

A time series is homogeneous when its statistical properties are not substantially altered by non-hydrological changes such as relocation, instrument replacement or changes in observation practice. A sudden shift in a rainfall record may represent a station move rather than climate variability.

Double-mass analysis compares cumulative values at a target station with cumulative values from a suitable regional reference. A change in slope suggests inconsistency, although the reference series itself must be checked. Statistical change-point tests can supplement graphical analysis but cannot identify the physical cause of a break without metadata.

Short gaps may be filled using nearby stations, interpolation or regression, provided the method is appropriate to the variable and time scale. Rainfall is spatially variable, so simple interpolation may be unsuitable for intense convective storms. Streamflow gaps may be estimated using correlated stations or rainfall–runoff models, but filled values should be marked and sensitivity-tested. Long gaps should not be disguised as observations.

2. Precipitation Analysis

2.1 Depth, intensity and duration

Precipitation depth P is usually measured in millimetres. Since 1 mm over 1 km² equals 1,000 m³, depth can be converted to volume by

V = PA

when P is expressed in metres and A in square metres. Rainfall intensity is depth per unit time, commonly mm/h. Design rainfall is described by depth, duration and frequency, often through intensity–duration–frequency (IDF) relationships.

For a storm with total depth P over duration t, the average intensity is

i = P/t.

This average does not describe short-duration peaks; a hyetograph is required when temporal distribution affects runoff generation.

2.2 Areal rainfall

A catchment rainfall estimate is needed for water-balance and flood studies. Common methods include the arithmetic mean, Thiessen polygons and isohyetal weighting. The arithmetic mean is suitable when gauges are relatively uniform in exposure and spatial distribution. Thiessen weighting assigns each station a polygon area, giving

P_A = \sum_{j=1}^{n} w_jP_j, \qquad \sum w_j=1,

where P_A is areal precipitation, Pj is the value at station j and wj is its area weight. Isohyetal analysis uses contours of equal rainfall and may better represent orographic patterns, but it requires sufficient data and professional judgement.

2.3 Temporal rainfall analysis

Rainfall series may be aggregated into annual, seasonal, monthly, daily or sub-daily totals. Aggregation reduces variability but can hide critical extremes. For flood design, annual maximum series or peaks-over-threshold data are often analysed. For water-supply planning, seasonal totals, cumulative deficits and low-flow periods may be more relevant.

A mass curve plots cumulative precipitation against time. Its slope represents rainfall rate, while departures from the long-term slope indicate wetter or drier periods. Double-mass curves and standardised anomalies help compare records across stations and periods.

3. Evaporation and Evapotranspiration

Evaporation is the transfer of water from liquid or solid surfaces to the atmosphere. Evapotranspiration (ET) includes evaporation from soil and water surfaces plus transpiration from vegetation. In catchment water balances, actual ET is constrained by available water, whereas reference or potential ET describes atmospheric demand under specified surface conditions.

The water balance for a catchment over a selected period may be written as

P = Q + ET + \Delta S + G + \epsilon,

where P is precipitation, Q is net runoff leaving the catchment, ET is evapotranspiration, ΔS is change in catchment storage, G represents net groundwater exchange where separately considered, and ε is the residual error. Sign conventions must be defined; for a closed catchment, G may be assumed negligible only after geological and groundwater conditions are assessed.

The Penman–Monteith approach is widely used for reference evapotranspiration. In daily form, a standard expression is

ET_0 = \frac{0.408\Delta(R_n-G)+\gamma\frac{900}{T+273}u_2(e_s-e_a)}{\Delta+\gamma(1+0.34u_2)},

where ET0 is reference evapotranspiration in mm/day, Δ is the slope of saturation vapour-pressure curve in kPa/°C, Rn is net radiation in MJ/m²/day, G is soil heat flux in MJ/m²/day, γ is the psychrometric constant in kPa/°C, T is mean air temperature in °C, u2 is wind speed at 2 m in m/s, es is saturation vapour pressure, and ea is actual vapour pressure, both in kPa. This equation requires compatible meteorological inputs and reference-surface assumptions. Missing or poor radiation, humidity or wind data can produce substantial error.

Pan evaporation may be converted to open-water evaporation using

E_w = K_p E_p,

where Ep is pan evaporation, Ew is estimated open-water evaporation and Kp is a site- and season-dependent pan coefficient. A universal coefficient should not be assumed.

4. Infiltration and Rainfall Excess

Infiltration is the entry of water into soil. Its rate depends on soil texture, initial moisture, vegetation, surface condition, rainfall intensity, compaction, crusting and preferential flow. Infiltration capacity generally declines during a storm as soil pores fill, although layered soils and macropores can produce complex behaviour.

A commonly used empirical representation is the Horton equation:

f(t)=f_c+(f_0-f_c)e^{-kt},

where f(t) is infiltration capacity, f0 is initial capacity, fc is final capacity, k is a decay coefficient and t is time from the beginning of rainfall under the calibration conditions. Actual infiltration cannot exceed available rainfall intensity; therefore the infiltration rate is generally the lesser of rainfall intensity and infiltration capacity. The equation is empirical and should not be extrapolated beyond the soil and storm conditions used for calibration.

The Green–Ampt model represents infiltration through a wetting front:

f = K_s\left(1+\frac{\psi_f\Delta\theta}{F}\right),

where f is infiltration capacity, Ks is saturated hydraulic conductivity, ψf is wetting-front suction head, Δθ is the change in volumetric water content across the front, and F is cumulative infiltration. This model assumes an idealised sharp wetting front and homogeneous soil. It can be useful for conceptual and event modelling, but parameter estimation is uncertain.

Rainfall excess is precipitation remaining after interception, depression storage, infiltration and other initial losses. It is the principal direct input to surface runoff. Loss models must be calibrated against observed storms or defensible soil and land-cover information.

5. Runoff, Streamflow and Hydrographs

5.1 Catchment water response

Runoff consists of surface runoff, subsurface stormflow and groundwater contribution. The relative importance of each component depends on geology, soil, slope, land cover, antecedent moisture and drainage network. Urbanisation generally increases impervious area, reduces infiltration opportunity and shortens travel time, although drainage design and storage can modify this response.

A streamflow hydrograph plots discharge against time. Its main features are baseflow, rising limb, peak discharge, recession limb, lag time, time to peak and event volume. A hyetograph plotted above the hydrograph helps assess the transformation of rainfall into runoff.

5.2 Hydrograph separation and indices

Hydrograph separation attempts to estimate baseflow and direct runoff. Straight-line, fixed-slope and recursive-filter methods are common, but none uniquely identifies physical flow pathways. Results are method-dependent and should be reported as estimates rather than direct observations.

Runoff volume is

V_Q = \int_{t_1}^{t_2} Q(t)\,dt.

For discrete observations with interval Δt, the trapezoidal approximation is

V_Q \approx \sum_{i=1}^{n-1}\frac{Q_i+Q_{i+1}}{2}\Delta t.

If Q is in m³/s and Δt is in seconds, VQ is in m³. Equivalent runoff depth over catchment area A is

P_Q = \frac{V_Q}{A}.

The event runoff coefficient is

C = \frac{P_Q}{P},

where P is event rainfall depth and both depths use the same units. C is not necessarily constant: it varies with antecedent wetness, storm intensity, season and urbanisation. Values above one indicate inconsistent event definitions, measurement error, storage effects or an invalid separation.

The unit hydrograph is the direct-runoff hydrograph resulting from one unit of effective rainfall uniformly distributed over a catchment during a specified duration, assuming linearity and time invariance. These assumptions often become weak for large storms, changing moisture conditions and urban catchments. Unit-hydrograph methods remain useful when their limitations are explicitly recognised.

5.3 Streamflow statistics

Important statistics include mean flow, median flow, coefficient of variation, flow-duration curves and low-flow indices. A flow-duration curve shows the percentage of time that a discharge is equalled or exceeded. High-flow portions support flood and channel studies; low-flow portions inform water-supply reliability and ecological assessment.

The coefficient of variation is

CV = s/\bar{x},

where s is sample standard deviation and x̄ is sample mean. It is dimensionless but becomes unstable when the mean is close to zero. Summary statistics should be accompanied by the record period, time step and treatment of missing values.

6. Probability Distributions and Frequency Analysis

6.1 Probability concepts

A random variable X represents a hydrological quantity such as annual maximum flood, annual rainfall or minimum seven-day flow. The cumulative distribution function is

F(x)=P(X\le x).

The exceedance probability is

P(X>x)=1-F(x),

for a continuous variable. Frequency analysis estimates quantiles, such as the discharge exceeded on average once in a specified recurrence interval.

Common distributions include the normal, lognormal, Gumbel (extreme-value type I), generalized extreme-value (GEV), gamma, Pearson type III and generalized Pareto distributions. Distribution selection should reflect the variable, sampling scheme, tail behaviour and diagnostic evidence; no distribution is universally correct.

6.2 Annual maxima and peaks over threshold

In an annual maximum series (AMS), the largest event in each year is retained. If the record contains n years, the plotting-position estimate of an event’s empirical non-exceedance probability may use a selected convention such as

F_i \approx \frac{i-a}{n+1-2a},

where i is the rank in ascending order and a is a plotting-position constant. Different conventions exist, so the chosen method must be stated.

A peaks-over-threshold (POT) series retains all independent events exceeding a threshold. POT can use more extreme events than AMS, but independence, threshold selection and separation of clustered storms require careful treatment. The generalized Pareto distribution is often used for exceedance magnitudes, while event occurrence may be modelled separately.

6.3 Parameter estimation and diagnostics

Distribution parameters may be estimated by moments, maximum likelihood or L-moments. Maximum likelihood can be efficient but sensitive to model specification and small samples. L-moments are often robust for extreme-value work and support regional analysis. Goodness-of-fit tests, probability plots, quantile plots and tail diagnostics should be used together. A formal test does not prove that a distribution is physically true; it only evaluates compatibility with the available sample.

For a Gumbel distribution, the quantile can be written

x_T = \mu - \beta\ln[-\ln(1-1/T)],

where μ is a location parameter, β is a positive scale parameter and T is the return period in years for an annual maximum interpretation. The equation assumes independent, identically distributed annual maxima and a stationary process. Parameter estimates and confidence intervals are essential because xT is estimated, not known exactly.

7. Return Periods and Design Interpretation

For an annual maximum event with annual exceedance probability p, the return period is

T=1/p.

Thus, a 100-year event has annual exceedance probability 0.01 under the conventional stationary interpretation. It does not mean that events occur regularly every 100 years. The probability of at least one exceedance during N independent years is

P_N=1-(1-p)^N=1-(1-1/T)^N.

For a 100-year event over 30 years, this probability is approximately 1−0.99³⁰ ≈ 0.260, or 26.0%.

Return periods can be misleading when observations are dependent, when the process is non-stationary, or when the series definition changes. A “100-year rainfall” also depends on duration and spatial averaging. Design communication should specify the variable, duration, area, time period, series type, stationarity assumption and confidence interval.

8. Uncertainty Assessment

8.1 Sources of uncertainty

Hydrological uncertainty has several components:

• measurement error from instruments and field procedures;
• sampling uncertainty caused by finite records;
• parameter uncertainty in fitted models;
• structural uncertainty caused by alternative model forms;
• input uncertainty in rainfall, soil, land use and climate data;
• natural variability or aleatory uncertainty;
• scenario uncertainty associated with future development or climate change.

These sources should not be collapsed into a single unexplained margin of error. Their relative importance depends on the decision and time horizon.

8.2 Confidence intervals and predictive intervals

A confidence interval describes uncertainty in an estimated population parameter or quantile under specified assumptions. A predictive interval describes uncertainty for a future observation and generally includes both parameter uncertainty and process variability. Confusing the two can lead to overconfidence in design values.

Bootstrap resampling can estimate uncertainty by repeatedly resampling observations and refitting the analysis. For dependent hydrological data, block bootstrap methods are preferable to ordinary independent resampling. Monte Carlo simulation propagates uncertain inputs and parameters through a model. Sensitivity analysis identifies which assumptions most affect the output.

8.3 Bayesian and regional approaches

Bayesian frequency analysis combines prior information with observed data to obtain a posterior distribution of parameters or design quantiles. It is useful where records are short but reliable regional or physical information exists. Priors must be justified and sensitivity to prior choices reported.

Regional frequency analysis pools information from hydrologically similar sites. It can improve estimation for short records, but heterogeneity in climate, catchment response and measurement quality must be assessed. Transferring information across unlike catchments can create a false impression of precision.

8.4 Non-stationarity and climate change

The stationarity assumption—that the probability distribution does not change with time—may be inappropriate where land use, reservoir operation, urbanisation or climate conditions are changing. Non-stationary models allow distribution parameters to depend on covariates such as time, temperature or catchment development. Such models require longer records, defensible covariates and careful validation. Apparent trends can also result from data inhomogeneity or serial dependence.

For adaptation planning, scenario analysis may be more informative than a single return-period estimate. Robust decisions can be explored using multiple climate projections, model structures and management options, rather than relying on one supposedly precise forecast.

9. Worked Example: Annual Maximum Flood Frequency

Suppose a gauging station has 20 years of independent annual maximum discharge data. A fitted stationary Gumbel model has parameters μ = 500 m³/s and β = 120 m³/s. Estimate the 50-year flood quantile.

For T = 50,

x50 = 500 − 120 ln[−ln(1−1/50)].

The probability term is 1−1/50 = 0.98. Then −ln(0.98) ≈ 0.020203, and ln(0.020203) ≈ −3.90197. Therefore,

x50 = 500 − 120(−3.90197)

≈ 500 + 468.24

≈ 968.2 m³/s.

The estimated 50-year flood is therefore approximately 968 m³/s, subject to parameter, model and sampling uncertainty. This is not a guarantee that a discharge of 968 m³/s will occur once in the next 50 years. Under the model, its annual exceedance probability is 2%, and the probability of at least one exceedance in 10 independent years is

1−(0.98)^10 ≈ 0.183,

or about 18.3%. In practice, confidence or credible intervals should accompany 968 m³/s, and the validity of stationarity and independence should be examined.

10. Illustrative Case: Flood Estimation in the Kalyan Catchment

Background

The fictional Kalyan Catchment is a 420 km² basin supplying a downstream town. It has 18 years of discharge records, three rainfall gauges and rapid urban growth in its lower reaches. A proposed flood-protection embankment requires an estimate of the design discharge.

Problem

The observed record contains missing stage data during two large storms. The rating curve was extended beyond the highest measured flow, and one rainfall station was relocated after eight years. A preliminary report fitted a Gumbel distribution directly to the recorded annual maxima without adjusting for these issues.

Analysis

The engineering team first preserved the raw series and reviewed station metadata. The rainfall record was tested for consistency using neighbouring gauges and a double-mass plot. The relocated station showed a slope change; its pre-relocation values were adjusted only after comparison with concurrent observations. Missing discharge values were reconstructed using a calibrated rainfall–runoff model, with a range of plausible parameters rather than a single deterministic estimate.

The team then compared AMS and POT analyses. Independence of POT events was checked by separating storms according to catchment response time. Several distributions were fitted using L-moments, and diagnostic plots were examined. The analysis also considered urbanisation by relating runoff coefficient and lag time to impervious-area growth. Rather than assuming stationarity, two scenarios were reported: current land use and projected development.

Approach and outcome

The final report presented design quantiles with confidence intervals, documented rating-curve extrapolation, and showed sensitivity to missing-flow reconstruction, distribution choice and future land use. The selected protection level was not based solely on the nominal 100-year discharge. It also considered consequences of overtopping, emergency storage, safe exceedance routes and the cost of raising the embankment in stages.

Learning points

1. Data quality and model choice can matter as much as the fitted distribution.

2. A longer apparent record is not necessarily more informative if it contains inconsistent observations.

3. Design decisions should consider consequences and uncertainty, not only a return period.

4. Scenario-based planning is appropriate where catchment conditions are changing.

5. Transparent documentation allows later analysts to update the estimate as new data become available.

11. Professional Practice, Governance and Technology

Hydrological data are public infrastructure when they support safety-critical decisions. Agencies should define responsibilities for station maintenance, metadata management, data release, quality flags and revision control. Ethical practice requires that analysts do not remove inconvenient observations, conceal uncertainty or present extrapolated values as direct measurements.

Digital telemetry, automated quality control, remote sensing, cloud databases and open data portals have improved timeliness and spatial coverage. Machine-learning methods can assist gap detection, rating-curve modelling, radar rainfall correction and streamflow forecasting. However, predictive performance outside the training range, data leakage, non-stationarity and lack of physical interpretability remain important concerns. AI should support—not replace—hydrological reasoning, field verification and uncertainty analysis.

Best practice includes:

• defining the decision and critical variable before selecting a method;
• maintaining raw, processed and modelled data separately;
• documenting units, timestamps, coordinate systems and transformations;
• checking physical plausibility and mass balance;
• validating models on independent periods where possible;
• reporting sensitivity, confidence or predictive intervals;
• testing alternative distributions and structural assumptions;
• communicating results in terms understandable to decision-makers;
• updating analyses when new observations or changed conditions arise.

Challenges and Limitations

Hydrological inference is constrained by short records, sparse networks, non-stationarity, spatial heterogeneity and uncertain extremes. Rainfall and streamflow observations may be serially correlated, and catchment responses may change after wildfire, urban development, deforestation or reservoir construction. Statistical models can describe observed data while failing under unprecedented conditions. Conversely, physically detailed models may contain many poorly known parameters.

No method eliminates uncertainty. The objective is to identify its sources, quantify the important components, test the robustness of decisions and communicate residual uncertainty honestly. The appropriate method is therefore determined not only by statistical fit but also by physical plausibility, data quality, consequence of failure and capacity for future updating.

Chapter Summary

Hydrological analysis transforms observations of precipitation, evaporation, infiltration and streamflow into information for water-resources decisions. Reliable analysis begins with well-designed monitoring, metadata, quality control, consistency checks and transparent treatment of missing values. Precipitation must be considered in terms of depth, intensity, duration and spatial averaging. Evapotranspiration, infiltration and runoff are governed by physical conditions and empirical models whose assumptions require calibration and evaluation.

Hydrographs reveal the timing and volume of catchment response, while flow-duration curves and runoff coefficients provide useful summaries. Frequency analysis uses probability distributions and either annual maxima or peaks-over-threshold data to estimate quantiles. Return period is a probability concept, not a schedule. Uncertainty arises from measurement, sampling, parameters, model structure, inputs, natural variability and changing future conditions. Confidence intervals, bootstrap methods, Monte Carlo simulation, Bayesian methods, regional analysis and scenario evaluation provide complementary tools.

Professional hydrology requires more than selecting a distribution. It requires defensible data, explicit assumptions, physical interpretation, sensitivity analysis, ethical reporting and decision-making that recognises the consequences of uncertainty.

Key Terms

Hydrological data: Observations describing water-cycle variables in space and time.

Rating curve: Relationship used to convert river stage to discharge.

Homogeneity: Stability of a record’s statistical properties with respect to non-hydrological changes.

Areal precipitation: Catchment-average precipitation estimated from point or gridded observations.

Hyetograph: Plot of precipitation intensity or depth against time.

Evapotranspiration: Combined evaporation and plant transpiration.

Infiltration: Entry of water into the soil surface.

Rainfall excess: Rainfall remaining after losses and available for direct runoff.

Baseflow: Streamflow sustained by delayed subsurface or groundwater contributions.

Hydrograph: Graph of discharge against time.

Runoff coefficient: Ratio of runoff depth to corresponding precipitation depth.

Flow-duration curve: Relationship between discharge and the percentage of time it is equalled or exceeded.

Annual maximum series: Series containing the largest observation from each year.

Peaks over threshold: Frequency-analysis method using independent events above a selected threshold.

Return period: Reciprocal of annual exceedance probability under specified assumptions.

Stationarity: Assumption that the probability distribution remains unchanged through time.

Parameter uncertainty: Uncertainty in estimated model or distribution parameters.

Structural uncertainty: Uncertainty arising from alternative model forms or assumptions.

Predictive interval: Range expressing uncertainty for a future observation or outcome.

Regional frequency analysis: Estimation method that pools information from hydrologically similar sites.

Review Questions

Short Answer Questions

1. Why is metadata essential when assessing hydrological records?

2. Distinguish precipitation depth from precipitation intensity.

3. What physical processes contribute to a streamflow hydrograph?

4. Define annual exceedance probability and return period.

5. Why should missing hydrological observations be flagged rather than silently replaced?

Long Answer Questions

1. Explain the principal sources of error in rainfall and streamflow measurement and describe suitable quality-control procedures.

2. Compare the Horton and Green–Ampt infiltration models, including their assumptions and limitations.

3. Explain how an annual maximum series differs from a peaks-over-threshold series.

4. Discuss the interpretation and limitations of the 100-year flood under stationary and non-stationary conditions.

5. Describe how parameter, model and input uncertainty can be assessed in a flood-frequency study.

Analytical / Application Questions

1. A catchment receives 85 mm of storm rainfall and produces 34 mm of direct runoff. Calculate the event runoff coefficient and discuss why the coefficient may differ for another storm.

2. A 50-year event has annual exceedance probability 0.02. Calculate the probability of at least one exceedance in 20 independent years and interpret the result.

3. A proposed dam relies on 15 years of flow data containing a rating-curve change and two years of major land-use change. Design an analysis plan that addresses data quality, frequency estimation, non-stationarity and uncertainty.

Multiple Choice Questions with Answers

Question 1

Which relation converts annual exceedance probability p to return period T under the conventional stationary interpretation?

A. T = p

B. T = 1/p

C. T = 1 − p

D. T = p²

Correct Answer: B. T = 1/p

Question 2

A 100-year event is exceeded in any one year with probability:

A. 0.1

B. 0.05

C. 0.01

D. 0.001

Correct Answer: C. 0.01

Question 3

Which instrument is commonly used to measure river stage before conversion to discharge?

A. Anemometer

B. Staff gauge

C. Pyranometer

D. Lysimeter

Correct Answer: B. Staff gauge

Question 4

If a catchment receives 80 mm of rainfall and produces 20 mm of direct runoff, its runoff coefficient is:

A. 0.05

B. 0.20

C. 0.25

D. 4.00

Correct Answer: C. 0.25

Question 5

Which method retains independent events exceeding a selected threshold?

A. Annual maximum series

B. Double-mass analysis

C. Peaks over threshold

D. Flow-duration analysis

Correct Answer: C. Peaks over threshold

Question 6

In the Penman–Monteith equation, which variable represents wind speed measured at 2 m?

A. Rn

B. u2

C. Δ

D. γ

Correct Answer: B. u2

Question 7

A flow-duration curve shows:

A. Rainfall intensity against storm duration

B. Discharge against percentage of time equalled or exceeded

C. Stage against channel width only

D. Evaporation against soil depth

Correct Answer: B. Discharge against percentage of time equalled or exceeded

Question 8

Which is an example of parameter uncertainty?

A. Uncertainty about whether a Gumbel or GEV model is appropriate

B. Uncertainty in the fitted scale parameter of a selected distribution

C. Uncertainty caused by future urbanisation scenarios

D. Random variation among individual storms

Correct Answer: B. Uncertainty in the fitted scale parameter of a selected distribution

Question 9

For a 100-year event over 30 independent years, the probability of at least one exceedance is approximately:

A. 1%

B. 10%

C. 26%

D. 100%

Correct Answer: C. 26%

Question 10

Why can a rating curve be unreliable at extreme floods?

A. Rainfall ceases during high flow

B. High flows are often extrapolated beyond direct measurements

C. Stage and discharge are always identical

D. Baseflow becomes zero in every river

Correct Answer: B. High flows are often extrapolated beyond direct measurements