C193a1 001205 this version not final
Comparing the RUNOFF kernel to laboratory experiments, including
super-critical laminar flows.
An
attempt is made to validate the runoff kernel algorithm in SWMM against earlier
findings on a laboratory-scale pavement. PCSWMM is used to calibrate the kernel
and generate plots that are compared to the experimental results obtained from
a laboratory rig described earlier. The
pavement was 100% impervious, with a 0.025 m/m slope over an area of 2.11 m2
and 1.171 m wide. A small time step of
1 s was used. It is shown that supercritical laminar flow cases are included.
INTRODUCTION
This part has been abstracted from the SWMM-RUNOFF
documentation (Huber et al.,
1988). Only the mathematics of overland
flow routing is reviewed - snowmelt, evaporation, infiltration, and groundwater
are not discussed in this chapter.
In SWMM-RUNOFF all subcatchments are assumed to be rectangular and each subcatchment is schematized so that three or four subareas (depending on whether snowmelt is simulated) are used to represent different surface properties. The slope of the idealized subcatchment is in the direction perpendicular to the width. Flow from each subarea moves directly to an outlet and does not pass over any other subarea. (Thus, it is not possible to route runoff from roofs over lawn surfaces, for instance). Subcatchments are subdivided into three subareas that simulate impervious areas with and without depression storage, and pervious area (with depression storage and infiltration). These are areas A1, A3, and A2 respectively in Figure 1. The width of the pervious subarea, A2, is the entire subcatchment width, whereas the widths of the two impervious subareas, A1, and A3 are in proportion to the ratio of their area to the total impervious area, as indicated in Figure 1. Of course, real subcatchments seldom exhibit the uniform rectangular geometries shown in Figure 1. In our case, however the laboratory pavement met these conditions, corresponding to subarea A3.
Figure
1 Subcatchment schematization. Each subarea flows instantaneously to an
outlet node. Flow from one subarea is
not routed over another. (Adapted
from Huber et al., 1988). (Note: Figures are appended in order at the
end of this paper)
Overland
flow is generated from each of the three subareas by approximating them as non‑linear
reservoirs, as sketched in Figure 2.
The width (and the slope and roughness) may generally be considered
calibration parameters, but this approach was not adopted in this study. Instead the initial
abstractions and Manning’s n were used.
The non‑linear
reservoir concept used in the kernel is established by coupling the continuity
equation with Manning's equation (Huber et al., 1988). Finally an equation for dd/dt is solved at
each time step by means of a simple finite difference scheme. For this purpose, the net inflow and outflow
on the right hand side of the equation must be averaged over the time
step. The rainfall excess i* is given
in the program as a time step average.
The average outflow is approximated by using the average between the old
and new depths.
Figure 2 Overland flow concept - depression storage is averaged over the area. (Adapted from
Huber et al., 1988).
If subscripts 1 and 2 denote the beginning and the end of a time step, respectively, the equation is:
[1]
where
d = water depth, m,
t = time step, s.
i* = rainfall excess,
=
rainfall/snowmelt intensity -
evaporation/infiltration rate, m/s
dp
= depth of depression storage, m
[2]
A = surface
area of subcatchment, m2,
W = subcatchment
width, m,
cm
= unit
conversion constant
= 1 for SI (m-s) units
= 1.49
for U.S. Customary (ft-s) units
n = Manning's roughness coefficient,
S = subcatchment slope, m/m.
The
subcatchment outflow rate Q (m3/s) is generated using Manning's
equation:
[3]
An
attempt was made to validate the above kernel against the findings of the
previous sections. Four tests were run and
plots were generated, calibrated, and compared to the experimental results
using a laboratory rig described earlier (James and Johanson, 1999; James and
Wylie, 1999; James et al., 1999).
The
input files were constructed based on the conditions applicable to overland
flow in the laboratory rig. The
pavement was 100% impervious, with a 0.025 m/m slope over an area of 0.000211
ha, and 1.171 m width. A small time
step of 1 s was used since the flow rates were expected to be small. Table 1 outlines the rainfall parameters
used in the simulation.
Each
test was run in PCSWMM98, then overlaid upon the observed hydrograph points of
Johanson (1967). Tests 1 and 2 required
calibration to better fit the data. The
difference was determined between observed flow rate and calculated flow rate,
which was then applied to the rainfall intensity input in RUNOFF. SWMM-RUNOFF was re-run and results compared
again. The resulting peak flow rates
differed by less than 1% in each run.
Table
1 Summary of SWMM-RUNOFF rainfall
parameters.
|
Parameter |
(Fig. 3) |
Test #2 (Fig. 4) |
Test #3 (Fig. 5) |
Test #4 (Fig. 6) |
|
Intensity (mm/h) |
117.5 |
120.3 |
a)
82.5 b)
121.2 |
a)
122.1 b)
83.1 |
|
Duration (s) |
120 |
60 |
a)
60 b)
60 |
a)
60 b)
60 |
Initial rainfall, a,
is followed by b, at given intensity and duration.
Figures
3 to 6 represent the results of the SWMM-RUNOFF runs for each test. Each figure represents the generated SWMM-RUNOFF
hydrographs superimposed on the original observed hydrographs of Johanson
(1967). It can be seen that SWMM-RUNOFF
calculates similar peaks, but different overall hydrograph shapes from those
observed. Thus, validation of the
program is not accurately attained.
Recall that the program does not account for induced upper surface shear
stresses due to rain, rain induced turbulence, transition to laminar flow,
induced lower surface shear stresses due to the transverse momentum of
infiltration, or transition between super- and sub-critical flow.
Figure 3 Long-duration comparison.
Figure 4 Short-duration comparison.
Figure 5 Variable increasing rainfall comparison.
Figure 6 Variable decreasing rainfall comparison.
A storage-routing model was
developed by Falk and Niemczynowicz (1979) and fitted to their observations of
rain on and runoff from thirteen small paved catchments in the city of Lund in
Sweden. They introduced a time lag t into the dynamic equation:
[4]
where:
S =
storage volume (mm)
Q =
outflow (mm×hour-1)
k, n, and t = model parameters
The process uses a step by
step procedure (Falk and Niemczynowicz, 1978).
To determine the model parameter t one must plot various relationships between runoff and storage for
different values of t and n. The values chosen are those that yield the least hysteresis
effect (where values of runoff are different for the rising and falling limb of
the hydrograph, for the same storage).
The continuity equation is:
[5]
where:
I =
rain intensity, mm×hour-1
To give the catchment water
budget for the time step (t-2t) to (t-t), equation 5 is changed to the discreet form:
[6]
Qt can be
determined by substituting the value of St-t (the only unknown variable in
equation 6) by the storage calculated in Equation 10. By expressing Q and I in mm×hour-1 and S in mm, the equation becomes:
[7]
Falk and Niemczynowicz
(1979) cite the following example:
Assume the following values
for parameters k and n have been found for a specific catchment:
k = 0.215
n = 0.667
Knowing the rainfall
intensities, the following steps will determine Qt :
Choose an appropriate time
step, t, say one minute.
Period 1 min:

Period 2 min:

Period 3 min:

The following equations can be used to
determine
and
:
[8]
[9]
The two storage terms, DS and St , can be determined using the following equations:
[10]
[11]
Table 2 shows calculations for
storage St for the first 5
of 20 minutes.
Table 2:
Storage and runoff calculations (first 5 minutes only).
|
Time |
Intensity |
I bar |
Qt |
Q bar |
delta S |
St |
|
min |
mm/hour |
mm/min |
mm/hour |
mm/min |
mm |
mm |
|
0 |
0 |
0 |
0.00 |
0.00 |
0.00 |
0.00 |
|
1 |
60 |
0.5 |
0.00 |
0.00 |
0.50 |
0.50 |
|
2 |
60 |
1 |
3.54 |
0.03 |
0.97 |
1.47 |
|
3 |
120 |
1.5 |
17.86 |
0.18 |
1.32 |
2.79 |
|
4 |
60 |
1.5 |
46.71 |
0.54 |
0.96 |
3.75 |
|
5 |
0 |
0.5 |
72.80 |
1.00 |
-0.50 |
3.26 |
|
|
|
|
|
|
|
|
Shown in Figure 7 is the rating curve, the plot of storage (mm) vs. outflow (mm×hour-1) for their example. The net rainfall used in this model is obtained from observed rainfall for the specified area by subtracting losses. In their catchment the water stored in depression storage is the only loss. They assume that only after the depression storage is filled can runoff occur - the model does not incorporate other losses such as evaporation or infiltration. By choosing Dt = t =1 minute, their method is evidently equivalent to using backward differences for storage, central differences for outflow, and t = 0.
Application to our experimental results:
We now apply the method to
the observations of Johanson (1967) that we have used above. Storage is computed in two ways: the first
is the same cumulative method as in the Falk and Niemczynowicz example, the
second uses equation 7 directly (k is obtained from the slope of the outflow
vs. storage plot, created using the storage determined by the cumulative
method). For the second method, a value
of n is chosen. The two storage values
calculated are compared by determining the percent difference, given by
equation 12.
[12]
The value of n is changed in
the second method, and the new computed storage is again compared with the
cumulative storage.
Figure
7 Rating curve for Falk and
Niemczynowicz example.
Figure 8 depicts the rating
curve for the cumulative storage vs. outflow for Johanson’s experimental data
shown in Figure 9. Table 3 compares the
two storage methods for the first 100 of 200 seconds. Results indicate that n=1 is better than n=0.667. The table also
facilitates depiction of the flow transitions during shallow overland flow on
pavements.
Figure
8 Rating curve for Johanson’s data.
The tables of computations
above allow further investigation of the concepts and effects of flow
transitions discussed in the previous sections. Both the Reynolds number and the
Froude number were calculated:
[13]
[14]
where
V =
runoff velocity (runoff flow/runoff area) m/s
d =
storage depth (m)
n
=
kinematic viscosity (m2/s)
g = gravitational acceleration
(m/s2)
Figures 9 and 10 show plots
of the Reynolds number and the Froude number for the rainfall input used in
Johanson’s data. In Figure 9, the
Reynolds number indicates laminar flow, while the Froude number indicates
sub-critical flow. This is expected due
to both the low volume and the low velocity of the runoff.
Table
3 Calculations of storage for
Johanson’s data (first 100 secs).
Input
|
Parameter n = 1
|
Parameter n = 0.667
|
||||||
|
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
9 |
|
0 |
0 |
0.00 |
0.00 |
0.00 |
0.00 |
0.00 |
0.00 |
0.00 |
|
5 |
120.78 |
0.69 |
0.08 |
0.01 |
7.64 |
0.08 |
0.01 |
7.56 |
|
10 |
120.78 |
0.69 |
0.25 |
0.01 |
24.32 |
0.25 |
0.01 |
24.23 |
|
15 |
120.78 |
0.69 |
0.42 |
0.01 |
41.00 |
0.42 |
0.01 |
40.91 |
|
20 |
120.78 |
20.44 |
0.57 |
0.20 |
36.56 |
0.57 |
0.07 |
49.52 |
|
25 |
120.78 |
61.33 |
0.68 |
0.61 |
6.77 |
0.68 |
0.16 |
52.53 |
|
30 |
120.78 |
91.63 |
0.74 |
0.92 |
17.38 |
0.74 |
0.20 |
53.90 |
|
35 |
120.78 |
102.92 |
0.78 |
1.03 |
25.40 |
0.78 |
0.22 |
55.52 |
|
40 |
120.78 |
108.55 |
0.80 |
1.09 |
28.95 |
0.80 |
0.23 |
56.81 |
|
45 |
120.78 |
102.21 |
0.82 |
1.02 |
20.47 |
0.82 |
0.22 |
59.85 |
|
50 |
0.00 |
79.65 |
0.78 |
0.80 |
2.15 |
0.78 |
0.19 |
58.96 |
|
55 |
0.00 |
55.67 |
0.68 |
0.56 |
12.43 |
0.68 |
0.15 |
53.50 |
|
60 |
0.00 |
39.46 |
0.61 |
0.39 |
22.03 |
0.61 |
0.12 |
49.89 |
|
65 |
0.00 |
29.60 |
0.57 |
0.30 |
27.10 |
0.57 |
0.10 |
47.12 |
|
70 |
0.00 |
25.37 |
0.53 |
0.25 |
27.52 |
0.53 |
0.09 |
44.24 |
|
75 |
0.00 |
22.56 |
0.50 |
0.23 |
27.00 |
0.50 |
0.08 |
41.56 |
|
80 |
0.00 |
17.62 |
0.47 |
0.18 |
29.15 |
0.47 |
0.07 |
39.99 |
|
85 |
0.00 |
15.50 |
0.44 |
0.16 |
28.97 |
0.44 |
0.06 |
38.24 |
|
90 |
0.00 |
13.39 |
0.42 |
0.13 |
29.07 |
0.42 |
0.06 |
36.82 |
|
95 |
0.00 |
11.27 |
0.41 |
0.11 |
29.48 |
0.41 |
0.05 |
35.72 |
|
100 |
0.00 |
11.98 |
0.39 |
0.12 |
27.15 |
0.39 |
0.05 |
33.89 |
1:
Time (s) 2:
Rainfall intensity (mm/hour)
3: Observed runoff (mm/hr) 4, 7: Accumulated storage (Falk and
Niemczynowicz method)
5,
8: Computed storage 6, 9 %
Difference between storage
Figure
9 Reynolds Number for Johanson’s data.
Figure
10 Plot Froude number for Johanson’s
data.
Further work done by Johanson
included varying the rainfall intensities during the same rainfall duration as
shown in Figures 5 and 6. Figures 11
and 12 show plots of the Reynolds and Froude numbers for the same increasing
and decreasing rainfall intensities.
Table 1 provides the intensities and the durations for test numbers 3
and 4.
Figure
11 Re and Fr for Johanson’s increasing
rain intensity data (test 3).
Figure 12 Re and Fr using Johanson’s decreasing rain
intensity data (test 4).
In Figures 11 and 12, the
Reynolds Number indicates laminar flow, whereas the Froude Number indicates
super-critical flow. Laminar
super-critical flow evidently is not commonly reported in the literature. It’s appearance here helps explain
conditions giving rise to the anomalous hump.
Recall that, for a given flow condition, laminar flow is more efficient
than turbulent flow. Thus the flow will
appear to accelerate at the cessation of rain, when the rain-induced turbulence
disappears. Calculated Froude numbers
are high, but even higher Froude numbers are obtained at the tail of the runoff
recession (not shown).
The rating curves for the varying rainfall
intensity data are plotted in Figures 13 and 14.
Figure
13 Rating curve for Johanson’s
increasing rain intensity data (test 3).
Figure 14 Rating curve for Johanson’s decreasing rain
intensity data (test 4).
How are these results to be interpreted? Thin sheet flow under
light rain on a steep impervious asphalt road is shown in Figure 15. The RUNOFF kernel does not explicitly include mathematics for most of the
processes evident in Figure 15, such as (1) induced shear stresses at the upper
surface due to the vertical transfer of rain momentum, (2) turbulence induced
by raindrop impacts, (3) flow transition to relatively efficient laminar flow,
(4) induced shear stresses at the lower surface due to the transverse momentum
of infiltration, (5) flow transition between super- and sub-critical flow. Flow
transitions are usually accompanied by
unstable depths, such as those seen in Figure 15.
Derived Reynolds Numbers were rather low, indicating laminar
flow, whereas the simultaneous Froude Numbers were rather high, indicating
super-critical flow. A literature search
was unsuccessful in finding other examples of laminar super-critical flow, which
evidently is not commonly reported.
In terms of the SWMM-RUNOFF rain-runoff kernel algorithm, however, no particular significance is imputed to these revelations, other than that the performance of the kernel nevertheless remains reasonable.
This chapter has described an investigation into the performance of the rain-runoff kernel algorithm in SWMM-RUNOFF. An input datafile was run to replicate earlier findings on a laboratory-scale pavement. PCSWMM is used to calibrate the kernel and generate plots that are compared to the experimental results obtained from a laboratory rig described earlier. The pavement was 100% impervious, with a 0.025 m/m slope over an area of 2.11 m2 and 1.171 m wide. A small time step of 1 s was used.
Results demonstrated that
SWMM-RUNOFF calculates similar peak flows, but different overall hydrograph
shapes than those observed on the laboratory rig. Validation of the program is not proven for the rising and
falling limbs of the hydrographs. However readers should recall that the program does not account for induced upper surface shear stresses due
to the vertical transfer of rain momentum, rain induced turbulence, flow
transition to relatively efficient laminar flow, induced lower surface shear
stresses due to the transverse momentum of infiltration, or transition between
super- and sub-critical flow, usually accompanied by unstable depths.
Further calculations were carried out to elucidate these flow transitions. A storage-lag method of routing was used to derive information for calculating transient Reynold’s and Froude numbers for the overland flow process. Derived Reynolds Numbers were rather low, indicating laminar flow, whereas the simultaneous Froude Numbers were rather high, indicating super-critical flow. A literature search was unsuccessful in finding other examples of laminar super-critical flow, which evidently is not commonly reported. In terms of the SWMM-RUNOFF rain-runoff kernel algorithm, however, no particular significance is imputed to these revelations, other than that the performance of the kernel nevertheless remains reasonable.
REFERENCES
Falk, J. and Niemczynowicz, J. 1979. Modelling of Runoff
From Impermeable Surfaces. Report No. 3024 for Dept. of Water Resources
Engn’rg, Lund Inst. of Tech., U. of Lund.
Huber, W.C., Dickinson, R.E, Roesner, L.A. and Aldrich,
J.A. 1988. Stormwater management model user’s manual, Version 4. EPA/600/3-88-001, US Environmental
Protection Agency, Athens, GA.
Izzard, C.F.
1944. The surface profile of
overland flow. Trans. Am.
Geophys. Union, 25:959-968.
Izzard, C.F.
1946. Hydraulics of runoff from
developed surfaces. Proc. Highway Research Board. 26:129-150.
Izzard, C.F. and Augustine, M.T. 1943. Prelim. report on
analysis of runoff from simulated rainfall on a paved plot. Trans. AGU.
2:500-509.
James, W. and Johanson, R.C. 1999. A note on an
inherent difficulty with the unit hydrograph method. Chapter 1, New Applications in Modelling Urban Water Systems
(monograph 7 in the series). CHI,
Guelph, ON. ISBN 0-9697422-9-0. pp.
1-21.
James, W. and Wylie, S.C. 1999. Numerical
techniques for overland flow from pavement.
Chapter. 5, Applied Modelling of Urban Water Systems (monograph 8 in the
series). CHI, Guelph, ON. ISBN
0-9683681-3-1. pp. 77-112.
James, W., Wylie, S.C. and Johanson, R.C. 1999.
A laboratory rig for testing runoff from paved surfaces. Chapter 6, Applied Modelling of Urban Water
Systems (monograph 8 in the series).
CHI, Guelph, ON. ISBN
0-9683681-3-1. pp. 113-122.
Johanson, R.C.
1967. System analysis of the
rainfall-runoff process. M.Sc.Eng. thesis, University of Natal, Durban, South
Africa.














