Professional Documents
Culture Documents
Hydrology 1B
Unit hydrograph
1.1. General hydrologic systems model
The amount of water stored in a hydrologic system, 5 may be related to the rates of inflow I and
outflow Q by the following integral equation of continuity:
𝑑𝑆
= 𝐼(𝑡) − 𝑄(𝑡) (1)
𝑑𝑡
If the water is stored in a hydrologic system, such as a reservoir (Figure 1), in which the amount of
storage rises and falls with time in response to I and Q and their rates of change with respect to time:
𝑑𝐼⁄𝑑𝑡 , 𝑑 2 𝐼⁄𝑑𝑡 2 , … , 𝑑𝑄⁄𝑑𝑡 , 𝑑 2 𝑄⁄𝑑𝑡 2 , …, The amount of storage at any time can be expressed by a
storage function as:
𝑑𝐼 𝑑 2 𝐼 𝑑𝑄 𝑑 2 𝑄
𝑆 = 𝑓 (𝐼, , 2 , … , 𝑄, , ,…) (2)
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡 2
The function f is determined by the nature of the hydrologic system being examined.
1
EECQ 4242: Eng. Hydrology 1B
𝑑𝐼 𝑑2𝐼 𝑑 𝑚−1 𝐼
+ 𝑏1 𝐼 + 𝑏2 + 𝑏3 2 + ⋯ + 𝑏𝑚 𝑚−1 (3)
𝑑𝑡 𝑑𝑡 𝑑𝑡
in which 𝑎1 , 𝑎2 , … , 𝑎2 , 𝑏1 , 𝑏2 , … , 𝑏𝑚 are constants and derivatives of higher order than those shown
are neglected. The constant coefficients also make the system time-invariant so that the way the
system processes input into output does not change with time.
Differentiating (3), substituting the result for 𝑑𝑆⁄𝑑𝑡 in (1), and rearranging:
𝑑𝑛 𝑄 𝑑 𝑛−1 𝑄 𝑑2𝑄 𝑑𝑄
𝑎𝑛 𝑛 + 𝑎𝑛−1 𝑛−1 + ⋯ + 𝑎2 2 + 𝑎1 +𝑄 =
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡
𝑑𝐼 𝑑2𝐼 𝑑 𝑚−1 𝐼 𝑑𝑚 𝐼
𝐼 − 𝑏1 − 𝑏2 2 − ⋯ − 𝑏𝑚−1 𝑚−1 − 𝑏𝑚 𝑚 (4)
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡
Equation (4) is the general hydrologic system model. It describes a lumped system because it contains
derivatives with respect to time alone and not spatial dimensions.
The linear reservoir model is a special case of equation (4) where 𝑎1 = 𝑘 and all other coefficients are
zero, thus:
𝑑𝑄
𝑘 +𝑄 = 1 (5)
𝑑𝑡
Equation (4) can be summarized as:
𝑁 ( 𝐷 ) 𝑄 = 𝑀 (𝐷 )𝐼 (6)
Where D = d⁄dt and N(D)and M(D)are the differential operators
𝑑𝑛 𝑑 𝑛−1 𝑑2 𝑑
𝑁 (𝐷 ) = 𝑎𝑛 + 𝑎𝑛−1 + ⋯ + 𝑎2 + 𝑎1 +1=
𝑑𝑡 𝑛 𝑑𝑡 𝑛−1 𝑑𝑡 2 𝑑𝑡
𝑑𝑚 𝑑 𝑚−1 𝑑2 𝑑
𝑀 (𝐷 ) = 𝑎𝑚 𝑚
+ 𝑏𝑚−1 𝑚−1
+ ⋯ + 𝑏2 2
+ 𝑏1 + 1
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑑𝑡
Solving equation (5) gives:
𝑀 (𝐷 )
𝑄 (𝑡 ) = 𝐼 (𝑡 ) (7)
𝑁 (𝐷 )
1.2.2. Response functions of linear systems
Solution of equation (7) follows two principles for linear system operations which are derived from
methods for solving linear differential equations with constant coefficients:
1. Principle of proportionality – If a solution 𝑓(𝑄) is multiplied by a constant c, the resulting
function 𝑐𝑓(𝑄) is also a solution
2. Principle of additivity or superposition – If two solutions 𝑓1 (𝑄) and 𝑓2 (𝑄) of the equation are
added, the resulting function 𝑓1 (𝑄) + 𝑓2 (𝑄) is also a solution of the equation
The solution adopted depends on the input function 𝑁(𝐷)𝐼, and on the specified initial conditions or
values of the output variables at 𝑡 = 0.
2
EECQ 4242: Eng. Hydrology 1B
Figure 2. Responses of a linear system to impulse inputs. (a) unit impulse response function (b) the
response to two impulses is found by summing the individual response functions
If the storage reservoir in Fig. 1 is initially empty, and then the reservoir is instantaneously filled with
a unit amount of water, the resulting outflow function 𝑄(𝑡) is the impulse response function.
Following the two principles of linear system operation cited above, if two impulses are applied, one
of 3 units at time 𝜏1 and the other of 2 units at time 𝜏2 , the response of the system will be:
3𝑢(𝑡 — 𝜏1 ) + 2𝑢(𝑡 — 𝜏2 ),
This is shown in in Fig. 7.2b.
Analogously, continuous input can be treated as a sum of infinitesimal impulses. The amount of input
entering the system between times 𝜏 𝑎𝑛𝑑 𝜏 + 𝑑𝜏 𝑖𝑠 𝐼(𝜏)𝑑𝜏.
Example: if
𝐼 (𝜏) = 𝑡ℎ𝑒 𝑝𝑟𝑒𝑐𝑖𝑝𝑖𝑡𝑎𝑡𝑖𝑜𝑛 𝑖𝑛𝑡𝑒𝑛𝑠𝑖𝑡𝑦, 𝑚𝑚⁄ℎ𝑜𝑢𝑟
3
EECQ 4242: Eng. Hydrology 1B
4
EECQ 4242: Eng. Hydrology 1B
5
EECQ 4242: Eng. Hydrology 1B
Figure 4. Response functions of a linear system. The response functions in (a), (b), and (c) are on a
continuous time domain and that in (d) on a discrete time domain
1
ℎ (𝑡 ) = [𝑔(𝑡) − 𝑔(𝑡 − ∆𝑡)] (11)
∆𝑡
𝑡 𝑡−∆𝑡
1
= [∫ 𝑢(𝑙 )𝑑𝑙 − ∫ 𝑢(𝑙 )𝑑𝑙 ]
∆𝑡
0 0
𝑡
1
= ∫ 𝑢(𝑙 )𝑑𝑙 (12)
∆𝑡
𝑡−∆𝑡
6
EECQ 4242: Eng. Hydrology 1B
As shown in Figure 4, 𝑔(𝑡) − 𝑔(𝑡 − ∆𝑡) represents the area under the impulse response function
between 𝑡 − ∆𝑡 and t, and h(t) represents the slope of the unit step response function g(t) between
these two time points.
Example. Determine the impulse, step, and pulse response functions of a linear reservoir with storage
constant 𝑘(𝑆 = 𝑘𝑄).
Solution:
The continuity equation (1) is
𝑑𝑆
= 𝐼(𝑡) − 𝑄(𝑡)
𝑑𝑡
The storage function of a linear reservoir is
𝑆 = 𝑘𝑄 ; after differentiating , get 𝑑𝑆⁄𝑑𝑡 = 𝑘 𝑑𝑄⁄𝑑𝑡
Combining the continuity and storage equations yields:
𝑑𝑄 𝑑𝑄 1 1
𝑘 = 𝐼 (𝑡) − (𝑄𝑡) 𝑜𝑟 + 𝑄(𝑡) = 𝐼(𝑡)
𝑑𝑡 𝑑𝑡 𝑘 𝑘
This a 1st order differential equation, and can be solved by multiplying bith sides of the equation bu
the integration factor 𝑒 𝑡⁄𝑘 :
𝑑𝑄 1 𝑡⁄𝑘 1
𝑒 𝑡 ⁄𝑘 + 𝑒 𝑄(𝑡) = 𝑒 𝑡⁄𝑘 𝐼(𝑡)
𝑑𝑡 𝑘 𝑘
𝑑 1
Simplifying, get: (𝑄𝑒 𝑡⁄𝑘 ) = 𝑒 𝑡⁄𝑘 𝐼 (𝑡)
𝑑𝑡 𝑘
Integrating from initial conditions 𝑄 = 𝑄0 𝑎𝑡 𝑡 = 0
𝑄𝑡 ,𝑡 𝑡
1
∫ 𝑑(𝑄𝑒 𝑡⁄𝑘 ) = ∫ 𝑒 𝜏⁄𝑘 𝐼(𝜏) 𝑑𝜏
𝑘
𝑄0 ,0 0
Comparing this equation with the convolution integral, the two equations are the same if 𝑄0 = 0
1 −(𝑡−𝜏)⁄𝑘
𝑇ℎ𝑒𝑟𝑒𝑓𝑜𝑟𝑒 𝑢(𝑡 − 𝜏) = 𝑒
𝑘
If 𝑙 is defined as the lag time 𝑡 − 𝜏, the 𝐢𝐦𝐩𝐮𝐥𝐬𝐞 𝐫𝐞𝐬𝐩𝐨𝐧𝐬𝐞 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧 of a linear reservoir is:
1 −(𝑙)⁄𝑘
𝑢 (𝑙 ) = 𝑒
𝑘
7
EECQ 4242: Eng. Hydrology 1B
The requirement that 𝑄0 = 0 implies that the system starts from rest when the convolution integral is
applied.
The unit step response is given by equation 10.
𝑡 𝑡
1 𝑡
𝑔(𝑡) = ∫ 𝑢(𝑙 )𝑑𝑙 = ∫ 𝑒 −(𝑙)⁄𝑘 𝑑𝑙 = [−𝑒 −𝑡⁄𝑘 ]0 = 𝟏 − 𝒆−𝒕⁄𝒌
𝑘
0 0
Figure 5. Response function of a linear reservoir with k = 3hour. Pulse response function is for a pulse
input of two hours duration
8
EECQ 4242: Eng. Hydrology 1B
Figure 6. A continuous time function Q(t), (a), can be defined on a discrete time domain either by a
sampled data system (b), in which instantaneous values of the continuous time function are used, or
by a pulse data system (c), in which the integral or average value of the function over the interval is
used.
The pulse data system is used for precipitation and the value of its discrete input function for the mth
time interval is:
𝑚∆𝑡
𝑃𝑚 = ∫ 𝐼 (𝜏)𝑑𝑡 𝑚 = 1, 2, 3, … (13)
(𝑚−1)∆𝑡
𝑃𝑚 is the depth of the precipitation falling during the time interval in units of length e.g. cm, mm etc.
9
EECQ 4242: Eng. Hydrology 1B
The sample data system is used for streamflow and direct runoff, so that the value of the system output
in the nth time interval (𝑡 − 𝑛∆𝑡) is:
𝑄𝑛 = 𝑄 (𝑛∆𝑡) 𝑛 = 1, 2, 3, … 14
𝑄𝑛 is the instantaneous value of the flow rate at the end of the nth time interval in units of discharge
e.g. 𝑚3 ⁄𝑠
This shows that input and output variables to a watershed system are recorded with different
dimensions and using different discrete data representations.
The effect of an input pulse of duration ∆𝑡 beginning at time (𝑚 − 1)∆𝑡 on the output at time 𝑡 = 𝑛∆𝑡
is measured by the value of the unit pulse response function:
ℎ[𝑡 − (𝑚 − 1)∆𝑡] = ℎ[𝑛∆𝑡 − (𝑚 − 1)∆𝑡] = ℎ[(𝑛 − 𝑚 + 𝐼 )∆𝑡],
given, following Equation (12), as
(𝑛−𝑚+1)∆𝑡
1
ℎ[(𝑛 − 𝑚 + 𝐼 )∆𝑡] = ∫ 𝑢(𝑙 )𝑑𝑙 (15)
∆𝑡
(𝑛−𝑚)∆𝑡
On a discrete time domain, the input function is a series of M pulses of constant rate:
𝑓𝑜𝑟 𝑝𝑢𝑙𝑠𝑒 𝑚, 𝐼 (𝜏) = 𝑃𝑚 ⁄∆𝑡 𝑓𝑜𝑟 (𝑚 − 1) ∆𝑡 ≤ 𝜏 ≤ 𝑚 ∆𝑡.
𝐼(𝜏) = 0 𝑓𝑜𝑟 τ > M∆t
In the case where the output is being calculated after all the input has ceased, i.e.
𝑎𝑡 𝑡 = 𝑛∆𝑡 > 𝑀∆𝑡
see Fig. 3b.
The contribution to the output of each of the M input pulses can be found by breaking the convolution
integral (Equation 8) at 𝑡 = 𝑛∆𝑡 into M parts:
𝑛∆𝑡
𝑄𝑛 = ∫ 𝐼(𝜏)𝑢(𝑛∆𝑡 − 𝜏)𝑑𝜏
0
∆𝑡 2∆𝑡
𝑃1 𝑃2
= ∫ 𝑢(𝑛∆𝑡 − 𝜏)𝑑𝜏 + ∫ 𝑢(𝑛∆𝑡 − 𝜏)𝑑𝜏 + ⋯ + (16)
∆𝑡 ∆𝑡
0 ∆𝑡
𝑚∆𝑡 𝑀∆𝑡
𝑃𝑚 𝑃𝑀
= ∫ 𝑢(𝑛∆𝑡 − 𝜏)𝑑𝜏 + ⋯ + ∫ 𝑢(𝑛∆𝑡 − 𝜏)𝑑𝜏
∆𝑡 ∆𝑡
(𝑚−1)∆𝑡 (𝑀−1)∆𝑡
where the terms 𝑃𝑚 ⁄∆𝑡 , 𝑚 = 1,2, . . . , 𝑀, are outside the integrals because they are constants.
In each of these integrals, the substitution 𝑙 = 𝑛∆𝑡 − 𝜏 is made so 𝑑𝜏 = − 𝑑𝑙;
the limit 𝜏 = (𝑚 − 1)∆𝜏 becomes 𝑙 = 𝑛∆𝑡 − (𝑚 − 1)∆𝑡 = (𝑛 − 𝑚 + 𝐼 )∆𝑡; 𝑎𝑛𝑑
the limit 𝜏 = 𝑚∆𝜏 becomes 𝑙 = (𝑛 − 𝑚)∆𝑡
10
EECQ 4242: Eng. Hydrology 1B
(𝑛−𝑚+1)∆𝑡
𝑃𝑚
= ∫ −𝑢(𝑙)𝑑𝑙 ) (17)
∆𝑡
(𝑛−𝑚)∆𝑡
= 𝑃𝑚 ℎ[(𝑛 − 𝑚 + 1)∆𝑡]
By substitution from equation 14.
After making these substitutions for each term in equation 16:
𝑄𝑛 = 𝑃1 ℎ[(𝑛∆𝑡] + 𝑃2 ℎ[(𝑛 − 1)∆𝑡] + ⋯
+𝑃𝑚 ℎ[(𝑛 − 𝑚 + 1)∆𝑡] + ⋯ (18)
+𝑃𝑀 ℎ[(𝑛 − 𝑀 + 1)∆𝑡]
Equation 18 is the convolution equation with input 𝑃𝑚 in pulses and output 𝑄𝑛 as a sample data
function of time.
𝑄𝑛 = ∑ 𝑃𝑚 𝑈𝑛−𝑚+1 (21)
𝑚=1
Combining equation(20) and (21) gives the discrete convolution equation for a linear system:
𝑛≤𝑀
𝑄𝑛 = ∑ 𝑃𝑚 𝑈𝑛−𝑚+1 (22)
𝑚=1
11
EECQ 4242: Eng. Hydrology 1B
The notation 𝑛 < 𝑀 as the upper limit of the summation shows that the terms are summed for 𝑚 =
1 , 2 , . . . , 𝑛 𝑓𝑜𝑟 𝑛 < 𝑀, but for 𝑛 > 𝑀, the summation is limited to 𝑚 = 1,2, . . . , 𝑀.
Example: suppose there are 𝑀 = 3 pulses of input: 𝑃1 , 𝑃2 , 𝑎𝑛𝑑 𝑃3 .
For the first time 𝑖𝑛𝑡𝑒𝑟𝑣𝑎𝑙 (𝑛 = 1), there is only one term in the convolution, that for 𝑚 = 1;
𝑄1 = 𝑃1 𝑈1−1+1 = 𝑃1 𝑈1
For 𝑛 = 2, there are two terms, corresponding to m = 1, 2:
𝑄2 = 𝑃1 𝑈2−1+1 + 𝑃2 𝑈2−2+1 = 𝑃1 𝑈2 + 𝑃2 𝑈1
For 𝑛 = 3, there are two terms, corresponding:
𝑄3 = 𝑃1 𝑈3−1+1 + 𝑃2 𝑈3−2+1 + 𝑃3 𝑈3−3+1 = 𝑃1 𝑈3 + 𝑃2 𝑈2 + 𝑃3 𝑈1
For 𝑛 = 4, 5, … there continues to be three terms:
𝑄𝑛 = 𝑃1 𝑈𝑛 + 𝑃2 𝑈𝑛−1 + 𝑃3 𝑈𝑛−2
The results of the calculation are shown diagrammatically in the following Figure 7. The sum of the
subscripts in each term on the right-hand side of the summation is always one greater than the
subscript of Q.
In the example shown in the diagram, there are 3 input pulses and 6 nonzero terms in the pulse
response function U, so there are 3 + 6 — 1 = 8 nonzero terms in the output function Q. The values
of the output for the final three periods are:
𝑄6 = 𝑃1 𝑈6 + 𝑃2 𝑈5 + 𝑃3 𝑈4
𝑄7 = 𝑃2 𝑈6 + 𝑃3 𝑈5
𝑄8 = 𝑃3 𝑈6
𝑄𝑛 and 𝑃𝑚 are expressed in different dimensions, and U has dimensions that are the ratio of the
dimensions of 𝑄𝑛 and 𝑃𝑚 to make (22) dimensionally consistent. For example, if 𝑃𝑚 is measured in mm
and 𝑄𝑛 in cumecs, then the dimensions of U are 𝑐𝑢𝑚𝑒𝑐𝑠⁄𝑚𝑚, which may be interpreted as cumecs
of output per mm of input.
12
EECQ 4242: Eng. Hydrology 1B
Figure 7. Application of the discrete convolution equation to the output from a linear system
13