Skip to content

Stream Flow, Hydrographs and Flood Frequency

CSIR NET Earth Sciences Part C Physical Geography: discharge measurement, Manning, rating curves, hydrographs, unit hydrograph, Gumbel and flood risk.

  • 21 min read
  • Updated
On this page · 14 sections
  1. Prerequisites
  2. The Hook: The “100-Year Flood” Twice in a Decade
  3. Measuring Stream Flow
  4. Runoff Generation
  5. The Hydrograph
  6. The Unit Hydrograph
  7. Flow Duration and Flood Frequency
  8. Memory Tricks
  9. Don’t Confuse
  10. Common Mistakes
  11. Practice
  12. Quick Revision Table
  13. Key Takeaways
  14. What’s Next

Prerequisites

The Hook: The “100-Year Flood” Twice in a Decade

Measuring Stream Flow

Discharge and continuity

Discharge \( Q \) (m\(^3\) s\(^{-1}\), “cumecs”) is the volume of water passing a cross-section per unit time:

$$\boxed{Q = A\,\bar{v}}$$

with \( A \) the cross-sectional area and \( \bar v \) the mean velocity. For steady flow without tributaries \( Q \) is constant along the channel (continuity), so where the channel narrows or shallows the velocity rises.

The velocity–area method

The channel is divided into vertical panels. In each, depth is sounded and velocity is measured with a current meter (cup or propeller type, or an acoustic Doppler profiler).

  • Velocity is lowest at the bed and highest just below the surface. In a vertical, the mean velocity is found at about 0.6 of the depth below the surface, or as the average of the velocities at 0.2 and 0.8 depth.
  • A floating object gives surface velocity; mean velocity is about 0.85 times the surface velocity (a commonly used coefficient, roughly 0.8–0.9).
  • Panel discharge = width × depth × mean velocity; total \( Q \) is their sum.

Other methods

  • Dilution (tracer) gauging. Inject a tracer (salt or dye) at a constant rate \( q \) with concentration \( C_1 \) into a stream of background concentration \( C_0 \). Downstream, after full mixing, the plateau concentration is \( C_2 \). A mass balance gives
$$Q = q\,\frac{C_1 - C_2}{C_2 - C_0}$$

Ideal for turbulent mountain streams where a current meter cannot be used.

  • Weirs and flumes. A structure of known shape makes discharge a fixed function of head (for a rectangular weir \( Q \propto H^{3/2} \)).
  • Slope–area method. After a flood, high-water marks give the water-surface slope and cross-section; Manning’s equation then gives the peak discharge.

Velocity from channel properties

Manning’s equation (SI units):

$$\boxed{\bar v = \frac{1}{n}\,R^{2/3}\,S^{1/2}}$$

where \( n \) is Manning’s roughness coefficient (about 0.02–0.03 for clean natural channels, higher for weedy or bouldery ones and floodplains). Chézy’s form is \( \bar v = C\sqrt{RS} \); the two are linked by \( C = R^{1/6}/n \).

The Froude number \( Fr = \bar v/\sqrt{gd} \) separates subcritical (\( Fr < 1 \), tranquil, most lowland rivers) from supercritical (\( Fr > 1 \), rapid, shooting) flow. A hydraulic jump marks the change from supercritical to subcritical. The Reynolds number \( Re = \bar v R/\nu \) is almost always far above the laminar limit: natural channel flow is turbulent.

The stage–discharge rating curve

Gauging stations record stage (water level) continuously, because it is cheap; discharge is measured only occasionally. The two are linked by a rating curve:

$$Q = a\,(h - h_0)^b$$

where \( h_0 \) is the stage of zero flow and \( b \) is typically between about 1.5 and 2.5 for natural sections. On log–log axes of \( Q \) against \( (h - h_0) \) the rating is a straight line of slope \( b \).

Ratings shift when the channel scours or fills, and show hysteresis (loop rating) during floods: for the same stage, discharge is higher on the rising limb than on the falling limb, because the water-surface slope is steeper as the flood wave arrives. Extrapolating a rating beyond the highest gauged flow is a major source of error in flood estimates.

Runoff Generation

MechanismWhen it happensWhere
Infiltration-excess (Hortonian) overland flowrainfall intensity exceeds infiltration capacitysemi-arid land, crusted, compacted or urban surfaces
Saturation-excess overland flow (Dunne and Black)soil is already saturated, so even gentle rain runs off; includes return flowvalley bottoms, hollows, near-channel zones in humid regions
Throughflow (subsurface stormflow)water moves laterally through soil, often via macropores and pipeshumid, vegetated hillslopes with permeable soils
Groundwater flow (baseflow)slow drainage from aquiferssustains flow between storms

The variable source area concept (Hewlett and Hibbert, 1967) holds that in humid catchments storm runoff comes from a saturated area that expands during a storm and contracts afterwards, mostly near channels. Horton’s model of uniform overland flow across a whole basin applies mainly where infiltration capacities are low.

The Hydrograph

A hydrograph plots discharge against time at a point. A storm hydrograph has:

PartMeaning
Rising limb (concentration curve)quickflow arriving from the basin
Peak (crest)maximum discharge
Falling limb (recession)drainage of channel and basin storage; its shape depends on basin characteristics, not on the storm
Baseflowgroundwater contribution beneath the storm response
Lag timetime from the centroid of effective rainfall to the peak (definitions vary; one also uses time from the start of rain)
Time baseduration of direct runoff
graph LR
    A["Rainfall"] --> B["Losses: interception, infiltration, depression storage"]
    A --> C["Effective rainfall"]
    C --> D["Direct runoff: overland flow + quick throughflow"]
    B --> E["Groundwater recharge"]
    E --> F["Baseflow"]
    D --> G["Storm hydrograph"]
    F --> G

Controls on hydrograph shape

FactorEffect on peak and lag
Basin shapecompact (circular) basins concentrate flow: higher peak, shorter lag; elongated basins give lower, longer peaks
Basin arealarger basin: larger total flow but lower peak per unit area and longer lag
Drainage densityhigh density (impermeable rocks, sparse vegetation) gives a fast, peaky response
Slopesteep slopes: faster runoff, sharper peak
Soil and rock permeabilitypermeable sandstone, chalk or karst: low peak, high baseflow
Antecedent moisturewet soils: more runoff, higher peak
Vegetationinterception and infiltration: lower, delayed peak
Urbanisationimpervious surfaces and storm drains: higher peak, shorter lag, steeper limbs, lower baseflow
Storm intensity and movementintense rain, or a storm moving downstream with the flow, gives a sharper, higher peak

Baseflow separation and recession

Simple separation methods draw a straight line from the start of the rise to a point on the recession \( N \) days after the peak. A widely used empirical rule is

$$N \approx 0.83\,A^{0.2} \quad (N \text{ in days},\; A \text{ in km}^2)$$

The recession is often exponential:

$$Q_t = Q_0\,k^t$$

where \( k \) is the recession constant (per day, less than 1). Plotting \( \log Q \) against time turns the recession into a straight line; breaks of slope separate quickflow, interflow and baseflow components.

The Unit Hydrograph

Definition and assumptions

The unit hydrograph (UH; Sherman, 1932) of duration \( D \) is the direct-runoff hydrograph produced by 1 cm (or 1 mm, or 1 inch) of effective rainfall falling uniformly over the basin at a uniform rate for \( D \) hours. It rests on three assumptions:

  1. Linearity (proportionality): a storm of \( P \) cm of the same duration produces ordinates \( P \) times the UH.
  2. Time invariance: the same storm gives the same response whenever it occurs.
  3. Superposition: the responses to successive rainfall blocks add, each shifted by its start time.

It also assumes effective rainfall is spatially uniform, which limits its use to basins of moderate size.

Deriving a UH from a recorded storm

  1. Separate baseflow to get the direct runoff hydrograph (DRH).
  2. Compute the runoff volume = sum of DRH ordinates × time interval.
  3. Divide by basin area to get the effective rainfall depth \( P_e \).
  4. Divide each DRH ordinate by \( P_e \) (in cm) to get the UH ordinates.

Check: the volume under a 1 cm UH must equal 1 cm over the basin area.

Using the UH: convolution

For a storm of several \( D \)-hour blocks, multiply the UH by each block’s depth, lag each product by the block’s start time, and add.

To change UH duration, build an S-curve (the response to continuous effective rain of 1 cm per \( D \) hours, made by summing UHs lagged by \( D \)), offset it by the new duration \( D' \), subtract and multiply by \( D/D' \).

Synthetic UHs (Snyder’s method, the SCS dimensionless UH) estimate the UH of an ungauged basin from its area, length and slope.

Flow Duration and Flood Frequency

Flow duration curve

A flow duration curve plots discharge against the percentage of time it is equalled or exceeded. A steep curve means a flashy river with little storage; a flat curve, a river sustained by groundwater, lakes or snowmelt. \( Q_{95} \) (flow exceeded 95% of the time) is a standard low-flow index.

Return period and probability

For an annual maximum series (the largest flood in each year), the return period \( T \) of a flood of magnitude \( x \) is the mean interval between years in which it is equalled or exceeded:

$$\boxed{T = \frac{1}{p}}$$

where \( p \) is the annual exceedance probability. A partial-duration (peaks-over-threshold) series uses every independent peak above a threshold; for frequent floods it gives somewhat shorter return periods than the annual series, and the two converge for large \( T \).

Plotting position (Weibull). Rank \( n \) annual peaks from largest (\( m = 1 \)) to smallest:

$$T = \frac{n+1}{m}, \qquad p = \frac{m}{n+1}$$

The bankfull discharge of many alluvial rivers has a recurrence interval of roughly 1–2 years on the annual series (about 1.5 years is the figure usually quoted), tying hydrology back to the Wolman–Miller argument.

Risk over a design life

The probability that a \( T \)-year flood is equalled or exceeded at least once in \( N \) years is

$$\boxed{R = 1 - \left(1 - \frac{1}{T}\right)^N}$$

When \( N = T \) and \( T \) is large, \( R \approx 1 - e^{-1} = 0.63 \): a 100-year flood has about a 63% chance of occurring in any 100-year period. Solving for \( T \) gives the design return period for an acceptable risk.

The Gumbel (Extreme Value Type I) method

Gumbel’s distribution of extremes is widely used for annual flood peaks. By the frequency-factor method,

$$\boxed{x_T = \bar x + K_T\,s}$$

where \( \bar x \) and \( s \) are the mean and standard deviation of the annual peaks, and for a long record

$$K_T = -\frac{\sqrt 6}{\pi}\left[0.5772 + \ln\ln\frac{T}{T-1}\right]$$
\( T \) (years)22.335102550100
\( K_T \)−0.1640.000.7191.3052.0442.5923.137
  • \( K_T = 0 \) at \( T = 2.33 \) years: in Gumbel’s distribution the mean annual flood has a return period of 2.33 years. The median flood (\( T = 2 \)) is smaller than the mean, because the distribution is skewed to the right.
  • For short records, \( K_T \) is corrected using the reduced mean and standard deviation for the sample size (Gumbel’s table); the values above are the large-sample limit.
  • Log-Pearson Type III, fitted to the logarithms of the peaks with a skew coefficient, is the US federal standard for flood frequency; the log-normal and generalised extreme value (GEV) distributions are also common.
  • Extrapolating far beyond the record length (for example a 1,000-year flood from 30 years of data) carries large uncertainty. Palaeoflood hydrology, which uses slackwater deposits and other flood evidence to extend the record by centuries to millennia, helps constrain the tail.

Memory Tricks

Don’t Confuse

ThisvsThat
Stage: water level (m)Discharge: volume per time (m\(^3\) s\(^{-1}\))
Infiltration-excess (Hortonian): rain too intenseSaturation-excess: soil already full
Lag time: rainfall centroid to peakTime base: whole duration of direct runoff
Unit hydrograph: 1 cm effective rain in D hoursS-curve: continuous effective rain
Annual maximum series: one peak per yearPartial-duration series: every peak above a threshold
Return period T: mean intervalAnnual exceedance probability: 1/T
Gumbel mean annual flood: T = 2.33 yrMedian annual flood: T = 2 yr
Subcritical flow: Fr < 1Supercritical flow: Fr > 1

Common Mistakes

Practice

All questions below are original practice questions written for this page. They are not from any CSIR NET paper.

Level 1: Foundation

  1. At what fraction of depth below the surface is the mean velocity in a vertical usually measured?
  2. What depth of effective rainfall defines a unit hydrograph (in the metric convention used here)?
  3. What is the annual exceedance probability of a 25-year flood?
Answers
  1. 0.6 of the depth (or the average of readings at 0.2 and 0.8 depth).
  2. 1 cm, falling uniformly over the basin in the stated duration.
  3. \( 1/25 = \) 0.04 (4% per year).

Level 2: Part B Style

Practice Q1. A salt solution of 20 g L\(^{-1}\) is injected at 0.02 m\(^3\) s\(^{-1}\) into a stream with negligible background salt. Downstream the steady concentration is 8 mg L\(^{-1}\). The stream discharge is about:

(a) 5 m\(^3\) s\(^{-1}\) (b) 20 m\(^3\) s\(^{-1}\) (c) 50 m\(^3\) s\(^{-1}\) (d) 500 m\(^3\) s\(^{-1}\)

Answer: (c) 50 m³ s⁻¹

Convert to the same units: \( C_1 = 20{,}000 \) mg L\(^{-1}\). \( Q = 0.02 \times (20{,}000 - 8)/(8 - 0) = 49.98 \approx 50 \) m\(^3\) s\(^{-1}\). The other options come from unit slips: convert both concentrations to the same unit before dividing.

Practice Q2. Compared with its pre-development state, a newly urbanised catchment will most likely show:

(a) a lower peak and longer lag (b) a higher peak, shorter lag and lower baseflow (c) a higher peak and higher baseflow (d) no change in peak but a longer time base

Answer: (b)

Impervious surfaces and drains speed runoff and reduce infiltration: peak rises, lag shortens and baseflow falls. (a) describes afforestation; (c) gets baseflow backwards.

Practice Q3. In a 19-year annual maximum series, the fourth-largest flood has a Weibull return period of:

(a) 4 years (b) 4.75 years (c) 5 years (d) 20 years

Answer: (c) 5 years

\( T = (n+1)/m = 20/4 = 5 \) years. (b) uses \( n/m \); (d) is the largest flood’s return period.

Practice Q4. A rating curve \( Q = a(h - h_0)^b \) with \( h_0 = 0.5 \) m passes through \( (h, Q) = (1.5 \text{ m}, 20 \text{ m}^3 \text{ s}^{-1}) \) and \( (3.0 \text{ m}, 110 \text{ m}^3 \text{ s}^{-1}) \). The exponent \( b \) is about:

(a) 1.5 (b) 1.86 (c) 2.46 (d) 5.5

Answer: (b) 1.86

\( b = \ln(110/20)/\ln(2.5/1.0) = \ln 5.5/\ln 2.5 = 1.705/0.916 = 1.86 \). Option (c) forgets to subtract \( h_0 \) (\( \ln 5.5/\ln 2 \)); (d) is the discharge ratio itself.

Level 3: Part C Style

Practice Q5. The 6-h unit hydrograph of a basin of 86.4 km\(^2\), at 3-h intervals, is 0, 5, 15, 22.5, 17.5, 11.25, 6.25, 2.5, 0 m\(^3\) s\(^{-1}\) cm\(^{-1}\). A storm gives 2 cm of effective rain in the first 6 h and 3 cm in the next 6 h. Which statements are correct?

  1. The peak direct runoff is 90 m\(^3\) s\(^{-1}\).
  2. The peak occurs 15 h after the storm began.
  3. The total direct runoff volume is \( 4.32\times10^6 \) m\(^3\).
  4. The peak equals 5 × 22.5 because the two blocks’ peaks coincide.

(a) 1, 2 and 3 only (b) 1 and 4 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 3 only

Multiply the UH by 2, and by 3 lagged by 6 h (two 3-h steps), then add:

t (h)036912151821242730
2 × UH01030453522.512.550
3 × UH (lagged)0154567.552.533.7518.757.50
DRH010306080906538.7518.757.50
  • 1 true: peak 90 m\(^3\) s\(^{-1}\).
  • 2 true: at 15 h.
  • 3 true: 5 cm over 86.4 km\(^2\) = \( 0.05 \times 86.4\times10^6 = 4.32\times10^6 \) m\(^3\) (check: sum of ordinates 400 × 10,800 s).
  • 4 false: 5 × 22.5 = 112.5; the peaks are 6 h apart, so they do not coincide.

Practice Q6. Annual flood peaks at a gauge have mean 1,200 m\(^3\) s\(^{-1}\) and standard deviation 400 m\(^3\) s\(^{-1}\) (use large-sample Gumbel). Which statements are correct?

  1. A flood of 2,000 m\(^3\) s\(^{-1}\) has a return period of about 24 years.
  2. The mean annual flood has a return period of 2.33 years.
  3. The chance that the 25-year flood is exceeded at least once in 10 years is about 34%.
  4. The 2-year flood is larger than the mean annual flood.

(a) 1, 2 and 3 only (b) 2 and 4 only (c) 1 and 2 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 3 only
  • 1 true: \( K = (2000 - 1200)/400 = 2.0 \). Solving \( 2.0 = -0.7797[0.5772 + \ln\ln(T/(T-1))] \) gives \( \ln\ln(T/(T-1)) = -3.142 \), \( T/(T-1) = e^{0.0432} = 1.0441 \), \( T \approx 23.7 \) years.
  • 2 true: \( K_T = 0 \) at \( T = 2.33 \).
  • 3 true: \( 1 - 0.96^{10} = 1 - 0.665 = 0.335 \).
  • 4 false: \( K_2 = -0.164 \), so \( x_2 = 1200 - 66 = 1{,}134 \) m\(^3\) s\(^{-1}\), smaller than the mean.

Practice Q7. After a storm, the recession of a river follows \( Q_t = 40 \times 0.9^t \) m\(^3\) s\(^{-1}\) (\( t \) in days). The basin area is 86.4 km\(^2\). Which statements are correct?

  1. Ten days after the peak, discharge is about 14 m\(^3\) s\(^{-1}\).
  2. Discharge falls to 10 m\(^3\) s\(^{-1}\) after about 13 days.
  3. By the rule \( N = 0.83A^{0.2} \), direct runoff ends about 2 days after the peak.
  4. A plot of \( Q \) (linear) against \( t \) is a straight line.

(a) 1, 2 and 3 only (b) 1 and 4 only (c) 2 and 3 only (d) 1, 2, 3 and 4

Answer: (a) 1, 2 and 3 only
  • 1 true: \( 40 \times 0.9^{10} = 40 \times 0.349 = 13.9 \).
  • 2 true: \( t = \ln(10/40)/\ln 0.9 = 13.2 \) days.
  • 3 true: \( 0.83 \times 86.4^{0.2} = 0.83 \times 2.44 = 2.0 \) days.
  • 4 false: an exponential recession is straight on a semi-log plot (log \( Q \) against \( t \)), not on linear axes.

Quick Revision Table

IdeaFact
Continuity\( Q = A\bar v \)
Mean velocity in a verticalat 0.6 depth, or mean of 0.2 and 0.8; surface velocity × about 0.85
Dilution gauging\( Q = q(C_1 - C_2)/(C_2 - C_0) \)
Manning\( \bar v = R^{2/3}S^{1/2}/n \); \( n \approx 0.02\text{–}0.03 \) clean channels
Chézy\( \bar v = C\sqrt{RS} \); \( C = R^{1/6}/n \)
Froude\( Fr = \bar v/\sqrt{gd} \); < 1 subcritical, > 1 supercritical
Rating curve\( Q = a(h - h_0)^b \), \( b \approx 1.5\text{–}2.5 \); loop rating in floods
RunoffHortonian (intensity > infiltration); saturation-excess; throughflow; variable source area
Hydrograph controlsshape, area, drainage density, slope, permeability, antecedent moisture, vegetation, urbanisation
Baseflow end\( N \approx 0.83A^{0.2} \) days (A in km\(^2\))
Recession\( Q_t = Q_0k^t \); straight on semi-log axes
Unit hydrograph1 cm, D hours, uniform; linearity, time invariance, superposition
UH derivationDRH / effective rain depth; check volume = 1 cm × area
Return period\( T = 1/p \); Weibull \( T = (n+1)/m \)
Risk\( R = 1 - (1 - 1/T)^N \); N = T gives about 63%
Gumbel\( x_T = \bar x + K_T s \); \( K_{2.33} = 0 \), \( K_{10} = 1.305 \), \( K_{50} = 2.592 \), \( K_{100} = 3.137 \)
Other distributionsLog-Pearson III, log-normal, GEV; palaeoflood data for the tail
Bankfull dischargerecurrence about 1–2 years (often quoted as about 1.5)

Key Takeaways

What’s Next