Overland Flow Time of Concentration on Very Flat Terrains

Transcription

1 Overland Flow Time of Concentration on Very Flat Terrains Ming-Han Li and Paramjit Chibber Two types of laboratory experiments were conducted to measure overland flow times on surfaces with very low slopes. One was a rainfall test using a mobile artificial rainfall simulator; the other was an impulse runoff test. Test plots were 6 ft (1.83 m) wide by 3 ft (9.14 m) long with slopes ranging from.24% to.48%. Surface types tested include bare clay, lawn (short grass), pasture (tall grass), asphalt, and concrete. A regression analysis was conducted to construct models for predicting flow times. Results predicted with regressed models were compared with those from empirical models in the literature. It was found that the slope variable in the regressed model from rainfall test data is less influential than that in existing models. Furthermore, the exponent for the slope variable in the regressed model for the impulse runoff condition is only 1/1th of those in existing models. Overall, most empirical models underestimate overland flow time for laboratory plots with very low slopes. The slope variable becomes insignificant in governing overland flow time when the slope is small. Antecedent soil moisture, not included in most empirical models, significantly affects time of concentration, which is included in the regressed models. Time of concentration is a primary basin parameter that represents the travel time from the hydraulically furthest point in a watershed to the outlet. The accuracy of estimation of peak discharge is sensitive to the accuracy of the estimated time of concentration (1). Therefore, the importance of an accurate estimate of the time of concentration cannot be overlooked. For empirical models that estimate time of concentration, one of the major variables computed is a nominal value of slope for the watershed. Previous literature (2 5) collected and listed most of the popular time-of-concentration formulas. All these formulas (6 22) are empirical models and are presented in Table 1. These models share the same general format formed by four variables: Length of the watershed, L; Surface roughness (usually Manning s roughness coefficient for overland flow), n; Slope of the watershed, S; and Rainfall intensity, i. M.-H. Li, Department of Landscape Architecture and Urban Planning, Texas A&M University and Texas Transportation Institute, 3137 TAMU, College Station, TX P. Chibber, PBS&J, 1211 Indian Creek Court, Beltsville, MD 275. Corresponding author: M.-H. Li, minghan@tamu.edu. Transportation Research Record: Journal of the Transportation Research Board, No. 26, Transportation Research Board of the National Academies, Washington, D.C., 28, pp DOI: /26-15 t c The general formula can be expressed as follows: a b y z = kl n S i () 1 where t c = time of concentration, k = constant, and a, b, y, and z = exponents. This equation exhibits a linear relationship for the logarithms of the variables involved. As indicated in Table 1, all models include the surface slope variable, S, raised to a negative power ranging from.2 to.5 that is, the slope variable is in the denominator and, therefore, the greater the surface slope, the less the time of concentration and vice versa. The problem occurs when such empirical models are to be used for estimating time of concentration on very flat terrains such as coastal plains. If existing empirical models are used, as the surface slope approaches zero the time of concentration approaches infinity. Engineering judgment suggests that such long times of concentration do not represent reality. METHODOLOGY Rainfall Test A rainfall test was conducted using a mobile artificial rainfall simulator (Figure 1) where test surfaces of low slope were prepared. The simulator included a hydroseeder with a 5-gal (1,89-liter) water tank and an 18-in.-high (.46-m) rain rack with spray nozzles mounted in a 5-ft (1.52-m) spacing. This rainfall simulator was designed to cover the entire test plot. The hydroseeder pump generated a flow ranging from 15 to 41 gal/min (.9 to.26 m 3 /s). With this range of flow rates, the precipitation rates generated by the spray nozzles ranged from 1.5 to 3.3 in./h (38.1 to 83.8 mm/h). Each test plot measured 6 ft (1.83 m) wide by 3 ft (9.14 m) long. This size allowed each rainfall test to run for more than 45 min using the maximum-capacity 5-gal water tank. A 45-min rainfall duration was used to ensure that each test would reach the time of concentration. Each rainfall test use the following procedure: Measure surface soil moisture (except on concrete and asphalt surfaces), Determine precipitation rate, Start hydroseeder pump, Observe the first flow of runoff at the outlet of the test plot, 133

2 134 Transportation Research Record 26 TABLE 1 Summary of Time of Concentration Models Publication and Year Equation for Time of Concentration (min) Remarks Williams (1922) (6) Kirpich (194) (7) Hathaway (1945) (8), Kerby (1959) (9) Izzard (1946) (1) Johnstone and Cross (1949) (11) California Culvert Practice (1955) (12) Henderson and Wooding (1964) (13) Morgali and Linsley (1965) (14), Aron and Erborge (1973) (15) FAA (197) (16) U.S. Soil Conservation Service (1975, 1986) (17, 18) Papadakis and Kazan (1986) (2) Chen and Wong (1993) (19), Wong (25) (2) TxDOT (1994) (21) Natural Resources Conservation Service (1997) (22) t c = 6LA.4 D 1 S.2 L = basin length, mi A = basin area, mi 2 D = diameter (mi) of a circular basin of area S = basin slope, % t c = KL.77 S y L = length of channel/ditch from headwater to outlet, ft S = average watershed slope, ft/ft For Tennessee, K =.78 and y =.385 For Pennsylvania, K =.13 and y =.5 t c =.8275 (LN).467 S.233 L = overland flow length, ft S = overland flow path slope, ft/ft N = flow retardance factor t c = 41.25(.7i + c)l.33 S.333 i.667 i = rainfall intensity, in./h c = retardance coefficient L = length of flow path, ft S = slope of flow path, ft/ft t c = 3L.5 S.5 L = basin length, mi S = basin slope, ft/mi t c = 6(11.9L 3 /H).385 L = length of longest watercourse, mi H = elevation difference between divide and outlet, ft If expressed as T c = kl a n b S y i z format: t c = KL.77 S.385 K = conversion constant t c =.94(Ln).6 S.3 i.4 L = length of overland flow, ft n = Manning s roughness coefficient S = overland flow plane slope, ft/ft i = rainfall intensity, in./h t c =.94L.6 n.6 S.3 i.4 L = length of overland flow, ft n = Manning roughness coefficient S = average overland slope, ft/ft i = rainfall intensity, in./h t c = 1.8(1.1 C)L.5 S.333 C = rational method runoff coefficient L = length of overland flow, ft S = surface slope, ft/ft t c = (1/6)Σ(L/V) L = length of flow path, ft V = average velocity in ft/s for various surfaces (The exponent of S, if converted from Manning s equation, will be.5) t c =.66L.5 n.52 S.31 i.38 L = length of flow path, ft n = roughness coefficient S = average slope of flow path, ft/ft i = rainfall intensity, in./h t c =.595(3.15).33k C.33 L.33(2 k) S.33 i.33(1+k) For water at 26 C C, k = constants (for smooth paved surfaces, C = 3, k =.5. For grass, C = 1, k = ) L = length of overland plane, m S = slope of overland plane, m/m i = net rainfall intensity, mm/h t c =.72(1.1 C)L.5 S.333 C = rational method runoff coefficient L = length of overland flow, m S = surface slope, m/m t c =.526[(1/CN) 9]L.8 S.5 CN = curve number L = flow length, ft S = average watershed slope, % The basin area should be smaller than 5 mi 2 (129.5 km 2 ). Developed for small drainage basins in Tennessee and Pennsylvania, with basin areas from 1 to 112 acres (.4 to 45.3 ha). Drainage basins with areas of less than 1 acres (4.5 ha) and slopes of less than.1. Hydraulically derived formula; values of c range from.7 for very smooth pavement to.12 for concrete pavement to.6 for dense turf. Developed for basins with areas between 25 and 1624 mi 2 (64.7 and km 2 ). Essentially the Kirpich (7) formula; developed for small mountainous basins in California. Based on kinematic wave theory for flow on an overland area. Overland flow equation from kinematic wave analysis of runoff from developed areas. Developed from airfield drainage data assembled by U.S. Corps of Engineers. Developed as a sum of individual travel times. V can be calculated using Manning s equation. Developed from USDA Agricultural Research Service data of 84 small rural watersheds from 22 states. Overland flow on test plots of 1 m wide by 25 m long. Slopes of 2% and 5%. Modified from FAA (16). For small rural watersheds. NOTE: 1 mi = 1.61 km; 1 ft =.348 m; 1 in. = 25.4 mm.

3 Li and Chibber 135 Weir for overflow Reservoir Runoff direction FIGURE 2 Reservoir used for impulse runoff test. Tested Surfaces and Number of Tests FIGURE 1 Rainfall test using artificial rainfall simulator. Monitor runoff rate continuously, Shut down hydroseeder pump after the runoff peaked for about 1 min, and Continue measuring runoff rate until runoff ceased. Impulse Runoff Test The impulse runoff test measured runoff travel time without artificial rainfall. Both pervious and impervious surfaces were tested. For pervious surface testing, test plots were kept saturated before testing. The notion was to minimize the effect of antecedent soil moisture and infiltration and focus on the effects of surface type, flow rate, and slope on travel time. Meanwhile, time of concentration was measured as the time for water to travel from the farthest point of the plot to the outlet. As with the rainfall test, researchers used a hydroseeder as the mobile water source. A large reservoir stored water filled by the hydroseeder. This reservoir, during an impulse runoff test, was placed on the upstream side of the test plot and as the reservoir capacity was exceeded, water overflowed the weir onto the plot (Figure 2). The hydroseeder pump generated a flow ranging from about 15 to 41 gal/min (.9 to.26 m 3 /s). The size of each test plot again measured 6 ft (1.83 m) wide by 3 ft (9.14 m) long, as in the rainfall test. Each impulse runoff test used the following procedure: Water pervious test plots, including bare clay and pasture, to create a saturated condition; Determine flow rate; Start hydroseeder pump to generate flow; Record time when water overtopped the weir; and Record travel time when water front reached the outlet. Five surface types were tested in this study. The details of each test type are described below: Bare clay (two plots). The texture for the soil tested was 21% sand, 31% silt, and 48% clay. The slopes of these two bare clay plots were.43% and.42%, respectively. Lawn (two plots). The lawn was established using common Bermudagrass (Cynodon dactylon). A grass height of less than 2 in. (5.1 mm) was maintained by mowing. The slopes of these two lawn plots were.48% and.24%, respectively. Pasture (two plots). The lawn was also formed using common Bermudagrass (C. dactylon). Grasses of 6 in. (.15 m) or taller were maintained to simulate pasture conditions. The slopes of these two pasture plots were.48% and.24%, respectively. Concrete (one plot). The plot was built on the taxiway of an old airport. The slope was.35%. Asphalt (one plot). An asphaltic cold mix was applied on top of an old airport taxiway to simulate the asphalt surface. The slope was.35%. The number of tests for different surfaces is presented in Table 2. The data and the range of that specific data category collected in this study are summarized in the following list: Slope:.24% to.48%; Rainfall intensity (for rainfall test): 1.5 to 3.3 in./h (38.1 to 83.8 mm/h); TABLE 2 Number of Tests Tested Surface Rainfall Test Impulse Runoff Test Bare clay Lawn 13 Not tested Pasture Concrete 1 5 Asphalt 7 16

4 136 Transportation Research Record 26 Runoff flow rate (for impulse runoff test): 15 to 41 gal/min (.9 to.26 m 3 /s); Antecedent soil moisture (bare clay, lawn, and pasture): 8% to 54%; Infiltration rate:.358 to.598 in./h (.91 to 1.52 mm/h); Soil texture: 21% sand, 31% silt, and 48% clay (bare clay, lawn, and pasture plots); and Raindrop size: to.7 in. ( to 1.8 mm) in diameter. RESULTS AND DISCUSSION Rainfall Test A typical hydrograph of the rainfall test observed in this study is shown in Figure 3. A rainfall test began at time zero when the artificial rainfall simulator was turned on. Initially, there was a no-flow period at the outlet. Time of beginning was recorded when the first flow was observed at the outlet. As the flow plateaued and fluctuated within 5% of the flow rate, the rate was considered as the peak, which determined the time to peak. Because time of concentration should involve only hydraulic travel time, the initial loss process (initial abstraction) was not considered as part of the time of concentration. Therefore, the time of concentration in the rainfall test was determined as the time to peak minus the time of beginning of runoff. The results of rainfall tests are presented in Table 3. A linear regression analysis on the rainfall test data was conducted to construct a model to predict the time of concentration. From a preliminary analysis, it was found that the antecedent soil moisture was negatively associated with the time of concentration (Figure 4). Therefore, the antecedent soil moisture variable, θ, was added to Equation 1 to create the new model as follows: a b x y z t = kl n θ S i ( 2) c where θ is antecedent soil moisture (percent) and x is an exponent of θ. In the new model, the watershed length L is 3 ft (9.14 m), which is the length of test plots. For the exponent of length L, the mean value,.5, of several models compiled by Papadakis and Kazan (2) was used. Similarly, regression analysis was conducted to construct models for the time of beginning, t b, and time to peak, t p, using the same initial model shown in Equation 1. The best-fit models for t c, t b, and t p are summarized in Table 4. Signs and values of exponents for each variable in all three models Q FIGURE 3 Time to Peak Time of Concentration Time of Beginning Typical hydrograph in a rainfall test. Time are generally similar. When compared with existing models, exponents of variables n, θ, and i are close to those of existing models and they are all statistically significant (P <.1). However, exponents of the slope variable, S, were found to be insignificant and their absolute values are less than those (.33) from existing models. It appears that S has the least influence on the time of concentration because the absolute values of the exponents for S are the least among all variables. This could result from the relatively few slopes tested in the study (.24%,.35%,.42%,.43%, and.48%). Or, if the slope data number is not the cause, the insignificant effect of S might indicate that when the slope is very small, the effect from variables n, θ, and i become dominant on the time of concentration. The final regressed model for time of concentration can be expressed as follows: t. 553L n θ S i ( 3) This model was compared with some existing models developed from both overland flow data [Kerby (9), Izzard (1), Henderson and Wooding (13), and Chen and Wong (19)] and watershed data [Papadakis and Kazan (2) and Natural Resources Conservation Service (22)]. Values used for variables in each model are presented in Table 5. The comparison result is plotted in Figure 5. It is apparent that all compared models underestimated the time of concentration for tested conditions. Note that Izzard s model predicted the time reasonably well with the majority underestimated and the minority overestimated. Impulse Runoff Test The results of impulse runoff tests are presented in Table 6. A linear regression analysis of the impulse runoff test data was also conducted to construct a model to predict the time of concentration. In this case, the time of concentration is the travel time from the farthest side of the test plot to the outlet. The initial regression model was modified from Equation 1 as follows: t c = c a b w y = kl n Q S ( 4) where t c = time of concentration (min), L = watershed length (ft) [3 ft (9.14 m) in this study], n = Manning s roughness coefficient for overland flow, Q = flow rate (gal/min), and a, b, w, and y = exponents. Again, the exponent (a) of variable L was set as.5, which is the mean value compiled by Papadakis and Kazan (2) from several existing models. The final model is presented in Table 7. Exponents of n and Q are statistically significant, whereas the exponent of S is insignificant, similar to the rainfall test result. Furthermore, the exponent for the slope variable in the regressed model for the impulse runoff condition is only one-tenth of those in existing models. Again, the slope variable becomes insignificant in governing overland flow time when the slope is small.

5 Li and Chibber 137 TABLE 3 Rainfall Test Results Time of Time to Antecedent Soil Rainfall Intensity Slope Surface Beginning (min) Peak (min) Moisture (%) (mm/h) (%) Bare Clay Bare Clay Bare Clay Bare Clay Bare Clay Bare Clay Bare Clay Bare Clay Bare Clay Bare Clay Bare Clay Lawn Lawn Lawn Lawn Lawn Lawn Lawn Lawn Lawn Lawn Lawn Lawn Lawn Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Concrete N.A Concrete N.A Concrete N.A Concrete N.A Concrete N.A Concrete N.A Concrete N.A Concrete N.A Concrete N.A Concrete N.A Asphalt N.A Asphalt N.A Asphalt N.A Asphalt N.A Asphalt N.A Asphalt N.A Asphalt N.A

6 138 Transportation Research Record 26 Ant. moisture content (%) R 2 = Time of conc. (min) (a) Ant. moisture content (%) R 2 = Time of conc. (min) (b) Ant. moisture content (%) 4 2 R 2 = Time of conc. (min) (c) FIGURE 4 Relationship between antecedent moisture content and time of concentration (rainfall tests): (a) bare clay, (b) lawn, and (c) pasture. TABLE 4 Regressed Models (Rainfall Test) N R 2 Constant b (for n) x (for θ) y (for S) z (for i) t c (.115) a (<.1) (<.1) (.192) (<.1) t b (.753) (<.1) (<.1) (.947) (.119) t p (.26) (<.1) (<.1) (.228) (<.1) a p-value in parentheses. TABLE 5 Variable Values Used for Model Comparison (Rainfall Test) Papadakis and Kazan/ Chen and Wong Henderson and NRCS a Kerby Model Izzard Model Model Wooding Models Model N (flow c (retardance C and k n(manning Curve Surface retardance factor) coefficient) (constant) roughness n) Number Bare clay.1.2 Not provided Lawn.3.46 C = 1, k = Pasture.4.6 Not provided.41 8 Concrete.2.12 C = 3, k = Asphalt.2.7 C = 3, k = a Natural Resources Conservation Service.

7 Li and Chibber TABLE 6 Impulse Runoff Test Results Observed Tc (min) Predicted T c (min) The observed times of concentration were also plotted against those predicted by the new and existing models (Figure 6). Values used for variables in each model are presented in Table 8. The Kirpich (7) and Natural Resources Conservation Service (22) models underestimated the observed times of concentration, the Texas Department of Transportation (21) overestimated them. The lower left cluster of data in Figure 6 indicates the test results on smooth surfaces (bare clay, asphalt, and concrete), and the higher clusters indicate the results on pasture. On the basis of comparison results, existing models cannot correctly predict the time of concentration on flat surfaces. Limitations The limitations associated with the laboratory experiment are summarized as follows: Data were collected only from laboratory experiments. Field studies are needed to verify the laboratory results. Only a few flat slopes were tested. The antecedent soil moisture had a significant effect on the time of concentration; however, changes in soil moisture during tests were not monitored. CONCLUSIONS This study Izzard Kerby Henderson-Wooding Chen-Wong Papadakis-Kazan NRCS FIGURE 5 Observed versus predicted time of concentration in the rainfall test by seven models: this study, Papadakis and Karan (2), Kerby (9), Izzard (1), Henderson and Wooding (13), Chen and Wong (19), and Natural Resources Conservation Service (NRCS) (22). Existing empirical models of estimating the time of concentration were mostly developed from data of common watersheds. The ability of these models to predict the time of concentration on very flat terrains has not been fully analyzed. This study used a controlled laboratory environment to measure times of concentration on test plots 6 ft (1.83 m) wide by 3 ft (9.14 m) long with slopes of less than.5%. Artificial rainfall tests and impulse runoff tests were TABLE 7 Travel Time Flow Rate Slope Surface (min) (m 3 /s) (%) Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Bare clay Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Pasture Concrete Concrete Concrete Concrete Concrete Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt Asphalt conducted. The results indicate that the slope variable was insignificant in governing overland flow time on low-slope surfaces. In addition, antecedent soil moisture was negatively associated with the time of concentration and was as influential as surface roughness, rainfall intensity, and flow rate to the time of concentration of low-slope surfaces. This finding suggests that soil moisture should Regressed Model (Impulse Runoff Test) N R 2 Constant b (for n) w (for Q) y (for S) t c a p-value in parentheses. (.185) a (<.1) (<.1) (.848)

8 14 Transportation Research Record 26 8 Observed Tc (min) This study Kirpich TxDOT NRCS Predicted T c (min) FIGURE 6 Observed versus predicted time of concentration in the impulse runoff test by four models: this study, Kirpich (7), Texas Department of Transportation (TxDOT) (21), and Natural Resources Conservation Service (NRCS) (22). TABLE 8 Variable Values Used for Model Comparison (Impulse Runoff Test) Kirpich Model TxDOT Model NRCS Mode Surface K C(runoff coefficient) Curve Number Bare clay Pasture Concrete Asphalt be considered in design and a moderate value should be chosen for risk balancing. The findings of this study indicate that a direction for future work would be through field-scale instrumentation and observation of a number of watersheds with low slopes. The rationale for performing the laboratory study was the lack of appropriate data as well as the desire to develop a relationship between the time of concentration and the explanatory variables without having to collect several years of data. In light of the realization that antecedent soil moisture can affect the time of concentration, some representative soil moisture content data would also be useful. ACKNOWLEDGMENT This study was sponsored by the Texas Department of Transportation. REFERENCES 1. McCuen, R. H., S. L. Wang, and W. J. Rawls. Estimating Urban Time of Concentration. Journal of Hydraulic Engineering, Vol. 11, No. 7, 1984, pp Papadakis, C. N., and M. N. Kazan. Time of Concentration in Small Rural Watersheds. Technical Report 11/8/86/CEE. Civil Engineering Department, University of Cincinnati, Cincinnati, Ohio, Goitom, T. G., and M. Baker, Jr. Evaluation of T c Methods in a Small Rural Watershed. In Channel Flow and Catchment Runoff: Proc. International Conference for Centennial of Manning s Formula and Kuichling s Rational Formula (B. C. Yen, ed.), University of Virginia, Charlottesville, 1989, pp Mays, L. W. Water Resources Engineering. John Wiley and Sons, New York, Computer Applications in Hydraulic Engineering, 5th ed. Haestad Methods, Inc., Watertown, Conn., Williams, G. B. Flood Discharges and the Dimensions of Spillways in India. Engineering (London), Vol. 134, 1922, p Kirpich, Z. P. Time of Concentration of Small Agricultural Watersheds. Civil Engineering, Vol. 1, No. 6, 194, p Hathaway, G. A. Design of Drainage Facilities. Transactions of the American Society of Civil Engineers, Vol. 11, 1945, p Kerby, W. S. Time of Concentration Studies. Civil Engineering, March, Izzard, C. F. Hydraulics of Runoff from Developed Surfaces. Proceedings of the Highway Research Board, Vol. 26, 1946, pp Johnstone, D., and W. P. Cross. Elements of Applied Hydrology. Ronald Press, New York, California Culvert Practice, 2nd ed. Department of Public Works, Division of Highways, Sacramento, Henderson, F. M., and R. A. Wooding. Overland Flow and Groundwater Flow from a Steady Rain of Finite Duration. Journal of Geophysical Research, Vol. 69, No. 8, 1964, pp Morgali, J. R., and R. K. Linsley. Computer Analysis of Overland Flow. Journal of the Hydraulics Division, Vol. 91, No. HY3, 1965, pp Aron, G., and C. E. Erborge. A Practical Feasibility Study of Flood Peak Abatement in Urban Areas. Report. U.S. Army Corps of Engineers, Sacramento, Calif., Circular on Airport Drainage. Report A/C B. Federal Aviation Administration, U.S. Department of Transportation, Washington, D.C., Urban Hydrology for Small Watersheds. Technical Release 55. U.S. Soil Conservation Service, Washington, D.C., Urban Hydrology for Small Watersheds. Technical Release 55, 2nd ed. U.S. Soil Conservation Service, Washington, D.C., Chen, C. N., and T. S. W. Wong. Critical Rainfall Duration for Maximum Discharge from Overland Plane. Journal of Hydraulic Engineering, Vol. 119, No. 9, 1993, pp Wong, T. S. W. Assessment of Time of Concentration Formulas for Overland Flow. Journal of Irrigation and Drainage Engineering, Vol. 131, No. 4, 25, pp Hydraulic Design Manual. Texas Department of Transportation, Austin, Ponds Planning, Design, Construction. Agriculture Handbook No. 59. U.S. Natural Resources Conservation Service, Washington, D.C., The Hydrology, Hydraulics, and Water Quality Committee sponsored publication of this paper.