Example of rainfall-runoff model: SCS curve number method

The curve number method is widely used for the determination of runoff in watersheds. It is a conceptual method for the prediction of runoff, based on empirical data of runoff from small catchments and hillslope plots monitored by USDA (Beven 2001). It predicts the volume of storm runoff after some initial retention before beginning of the runoff. It responds to major parameters important for runoff production within a watershed

44 2.4 Modeling rainfall-runoff processes in semi-arid regions

including soil type, land use or treatment and surface condition. The SCS curve number method assigns for each part of a watershed the curve number based on the land use, antecedent moisture condition and hydrologic soil group. The curve number presents a specific relationship between storm rainfall and percentage of runoff (Fig. 2.16).

The SCS curve number method was successfully applied in the dry areas of Yemen (Farquaharson et. al. 1997) and Australia (Dilshad & Peel 1996). Colombo & Sarfatti (1997) estimated runoff within two subcatchments of the Marib river in Eritrea by choosing the appropriate runoff curve number through analysis of remotely-sensed data.

(Landsat TM and SPOT). Abu-Awwad & Shatanawi (1997) used the SCS method for calculation of runoff volumes of the Wadi Bhutum (Jordan) (140 km²) and a rainfall amount of 100 to 150 mm/a. They used rainfall equivalences of return periods of 2,5 and 10 years. Farquaharson et al. (1997) pointed out that especially with low rainfall, runoff coefficient are mainly dependent on watershed characteristics such as soil cover or slope.

Fig. 2.16 Relationship between storm rainfall and percentage runoff predicted by the USDA curve number method (Beven 2001)

Drake et al. (1999) used the SCS curve number method in the context of modelling soil erosion at global and regional scales using remote sensing techniqes and GIS techniques.

The SCS curve number method has been incorporated into various rainfall-runoff models such as CREAMS (Chemicals, Runoff and Erosion from Agricultural Management Systems model), WEPP (Water Erosion Prediction Project) and the Soil Water Assessment Tool (SWAT) (Beven 2001). Ponce & Hawkins (1996) pointed out that advantages of the method were is simplicity and predictability. A major disadvantage is the unclear definition of the antecedent moisture condition. Baten (1994) used the curve

numbers to define the role of land cover in the runoff generation process within a version of knowledge processing for watershed flood runoff simulation in an object-oriented data environment. Mattikalli et al. (1996) used the US SCS classification of land use and soils within an integrated geographical information system including a runoff and water quality model. The runoff is estimated in terms of volume of precipitaton and potential maximum storage or retention. The SCS method is based on the empirical assumption that the ratio of actual to initial runoff is equal to the ratio of actual retention ot the potential retention.

Q P I

S = potential maximum retention after runoff begins Ia = initial abstraction ratio

P = rainfall [mm]

Considering purely physics this relation is not justified (Beven 2001). This method is a good representation of the influence of the Hortonian concept. During further modifications a factor was added to the potential retention. Ia, the initial abstraction, is the factor responsable for interception, initial infiltration, surface depression storage and evapotranspiration. It is related to retention by a linear dependance:

I_a = λ [2.4]S

λ = initial abstraction ratio

S = potential maximum retention after runoff begins [mm]

Ia = initial abstraction [mm]

The overland flow in a rainfall event or the effective rainfall is then computed as follows:

Q P S

S = potential maximum retention after runoff begins [mm]

λ = initial abstraction ratio

The initial abstraction ratio is mostly set to the empirical value of 0.2. Ponce &

Hawkins (1996) critized this adoption of the initial abstraction ratio to 0.2 in their review of the SCS curve number method.They referred to other studies showing variing values of

46 2.4 Modeling rainfall-runoff processes in semi-arid regions

the initial abstraction ratio. If P is lower than Ia the runoff is nil. The maximum potential soil water retention is then computed according to:

S = (1000CN − ) .∗ 10 2 54

[2.6]

CN = curve number

S = storage volume [mm]

CN is the curve number which can be taken from published data and is a dimensionless number. Popularity of the method mainly comes from the dimensionless number CN which is tabulated for various soil types and conditions (SCS 1972; USDA-SCS 1985). To account for the spatial variability of soil moisture, the original USDA procedure adjusts the CN value on the total rainfall of 5 preceeding days. Therefore values for wet humid and dry conditions are suggested. Further studies showed that this adjustment is ambiguous since the dependance on antecedent soil moisture is impossible to be defined on a function (Sommer 1997). Other approaches realised within the CREAMS model adjust the storage factor by using a ration of actual to saturated water content. Under dry soil moisture condition the factor of storage is only dependant on the maximum infiltration rate (Sommer 1997).

Wood & Blackburn (1984) found out that the antecedent moisture is an important factor using the SCS curve number method on data of 1600 runoff plots in Nevada, Texas and New Mexico and found big differences between computed and observed values. At dry conditions the method slightly underestimate the runoff, but under wet conditions the runoff is strongly overpredicted. They also found out that the results were better on plots with low vegetation cover approaching bare conditions under crop cover than on plots with denser vegetation cover. Low vegetation cover approaches the conditions of soil used for cultivation of crops. Wood & Blackburn (1984) pointed out that the crust formation in soils under semi-arid climate often deludes the results: A sandy loam with an impeding crust, classified as hydrological group B had 60.5% more runoff than a soil classified as hydrological group D. Therefore the hydrological soil group definition should be modified according to rangeland conditions (Table 2.10).

Table 2.10 Possible modifications of the hydrologic soil group definition for SCS procedure in rangelands (Wood & Blackburn 1984) (A – D: Hydrological soil group)

Hydrological soil group

Description

A Sandy texture or well-aggregated granular structure with vesicular pores B Massive or weak platy structure with few to common vesicular pores

C Moderate platy structure with common vesicular pores, massive with many vesicular pores, or clayey and weakly structured

D Strong platy structure with many vesicular pores or clayey and massive

If more information on soil crusts would be available for the different soils within the project area, the differentiation into the hydrological soil groups could be improved.

Table 2.11 Influence of antecedent soil moisture (AMS) on curve numbers

5 day antecedent rainfall (mm) Curve Number

Growing season Dormant season

CN1 (AMS I) < 36 < 13

CN2 (AMS II) 36 - 53 37 - 48

CN3 (AMS III) > 53 > 28

Farquaharson et al. (1997) adjusted the CN value dynamically over the year according to antecendent precipitation ratios. They calibrated their model on the rainfall and runoff records of six wadis in Yemen over a period of about five years. The runoff coefficient did no fall below 5%.

Since the hypothesis of the antecedent soil moisture is not very clear. Mishra et al.

(2003) developed a modified SCS-CN-Method based accounting for the static portion of infiltration and the antecedent moisture. They included the soil characteristics by soil-moisture content and the hydraulic conductivity-soil-moisture content. They postulate that the modified method approximates more the physically based infiltration process.

Nevertheless the method has still to be more tested. On the contrary to the original method it requires additionally the prior knowledge of the minimum infiltration rate.

The runoff is mainly dependent on the basin characteristics and the antecedent rainfall conditions. This is especially true since the magnitude of rainfall storms is independent on

48 2.4 Modeling rainfall-runoff processes in semi-arid regions

the average rainfall. Another approach could be the composite one, using an average CN number for the project region. The results of the distributed approach are higher than the composite approach (Grove et al. 1998) who investigated different watersheds. Grove et al.

(1998) who investigated different watersheds showed that with higher curve numbers the difference between composite and distributed CN gets smaller.

Some authors tried to combine the SCS method with other parameters such as slope length (Sommer 1997). Sommer (1997) introduced a new factor to the calculation of the rainfall-runoff curve number: the slope length. Since the rainfall-runoff curve number do not take into account transmission losses, slope length may be one of the parameters accounting for these losses. He introduced a formula according

Q e P T e CN

T = threshold for runoff generation P = precipitation [mm]

Q = runoff [mm]

He calibrated the formula according to measured data within rainfall-runoff plots situated at Tel Hadya, Syria. The constants were determined by regression analysis of the observed data. Since the calibration with runoff data is needed this approach could be only used in areas where sufficient data on runoff amounts exist.

2.5 Application of remote sensing in hydrology