Application of Natural Resources Conservation Service – Curve Number Method for RunoffEstimation with GIS in the Kardeh Watershed, Iran 5761. IntroductionDue to serious soil erosion and water deficiency in most areas of Iran, natural resources conservation isa vital issue. Despite these pressing issues, few studies have been carried out in rainfall-runoffmodelling in Iran. Conventional methods of runoff measurement are costly, time consuming, error-prone and difficult because of inaccessible terrain in many of the watersheds. Thus, the use of newtools, for instance GIS, to generate supporting land-based data for conserving soil and water resourcesin watershed planning is very much needed. In addition, most basins in Iran do not have sufficientnumbers of gauges to record rainfall and runoff. Scarcity of reliable recorded data therefore is anotherserious problem which planners and researchers face for the analysis of the hydrology of arid regions. There are several approaches to estimate ungauged basin runoff. Examples are the University ofBritish Columbia Watershed Model (UBCWM), Artificial Neural Network (ANN), SCS CurveNumber model, and Geomorphological Instantaneous Unit Hydrograph (GIUH). Among thesemethods, the SCS method (now called Natural Resources Conservation service Curve Number method(NRCS-CN)) is widely used because of its flexibility and simplicity. The method combines thewatershed parameters and climatic factors in one entity called the Curve Number (CN). Manyresearchers (Pandey and Sahu, 2002; Nayak and Jaiswal, 2003; Zhan and Huang, 2004; Gandini andUsunoff, 2004) have utilized the Geographic Information System (GIS) technique to estimate runoffCurve Number value throughout the world. In India, Pandy and Sahu (2002) pointed out that the landuse/land cover is an important parameter input of the SCS-CN model. Nayak and Jaiswal (2003) foundthat there was a good correlation between the measured and estimated runoff depth using GIS and CN.They concluded that GIS is an efficient tool for the preparation of most of the input data required bythe SCS curve number model. Akhondi (2001) pointed out that correlation between observed andestimated discharge using CN method is decreased by increasing watershed area. While having runoffdata is essential in all watershed development and management plans, very little work has beenpreviously done in the watersheds of Iran in estimating runoff from rainfall in ungauged watersheds.This study emphasizes the use of GIS technique to develop a data base containing all the informationof the study watershed for direct runoff depth estimation using the NRCS-CN model. The objective ofthis study was to evaluate the use of NRCS Curve Number method with GIS for estimating runoffdepth and the effect of slope on CN values and runoff generation in a well-equipped gauged watershed.If the estimated runoff data are accurate compared to observed runoff in this watershed, then theNRCS-CN method can be recommended for estimating runoff in ungauged watersheds of the region.2. Materials and methods2.1. Study areaThis study was conducted in the Kardeh watershed about 42 km north of Mashhad, Khorasan Razaviprovince, north eastern Iran (Figure 1). The watershed lies between 59º 26´ 3˝ to 59º 37´ 17˝ Elongitude and 36º 37´ 17˝ to 36º 58´ 25˝ N latitude. Kardeh watershed is 448.2 km2 in size. Theelevation of the watershed ranges from 1320 to 2960 m above mean sea level. The climate of thewatershed is semi-arid. The mean annual precipitation is about 296.4 mm. Temperature in the studywatershed is low with a mean annual of 11.6 °C. The mean relative humidity is approximately 52.6%,but varies from 32.1 % in August to 82.3% in February. In most parts of the study watershed, topsoil is loamy and the subsoil is sandy clay loamyexcept in alluvial deposits that have relatively heavy texture of clay. In barren areas where soil isshallow, fine platy structure surface soil and compressed blocky structure subsurface soil are be sound.The major land uses in the study watershed are shown in Figure 1. About 73% of Kardeh has occupiedby rangelands. Forested areas have been degraded because of severe utilization over the time and onlysingle trees are a sign of degraded forest in highlands (Table 1).
577 Mahboubeh Ebrahimian, Lai Food See, Mohd Hasmadi Ismail and Ismail Abdul MalekTable 1: Land use/ cover classes present in the Kardeh catchment* Land use/ Land cover Area (km²) % of total area Dry farmland (rainfed farming) 66.90 15 Forest Thin 25.50 5.7 Fair 5.20 1.2 Rangeland Good condition 32.80 7.3 Fair condition 92.70 20.7 Poor condition 204.40 45.5 Orchards and irrigation farmland 17.40 3.9 Settlement 0.28 0.1 Rocks 2.90 0.6 Total area 448.20 100*Watershed Management Department of Khorasan Razavi Jihad-e-Agriculture Organization, 1996 Figure 1: Location of the study area in Iran.2.2. Data sourcesTopographic maps at the scale of 1:25000 prepared by the National Cartographic Centre (NCC) wereused for demarcation of study watershed border and as basic maps in generation of other ancillarymaps. Digitized land use/cover map prepared by the Watershed Management Department of KhorasanRazavi Jihad-e-Agriculture Organization in 1996 was used for identification of types and areas of landuse. Digitized soil map at the scale of 1:25000 and soil information were taken from Khorasan RazaviGeneral Office of Natural Resources. Meteorological data including rainfall, evaporation, andtemperature and runoff data (1990-2000) were obtained from Khorasan Razavi Regional WaterAuthority.
Application of Natural Resources Conservation Service – Curve Number Method for RunoffEstimation with GIS in the Kardeh Watershed, Iran 5782.3. Software used for data processingArc View version 3.3 powerful Geographic Information System (GIS) software was used for creating,managing and generation of different layers and maps. The SPSS software was used for statisticalanalysis of data. The Microsoft Excel was used for mathematical calculations. Figure 2: Land use/ cover map of the study watershed.2.4. Generating hydrologic soil group (HSGs) mapThe hydrologic soil group is an attribute of the soil mapping unit (each soil mapping unit is assigned aparticular hydrologic soil group: A, B, C, or D). In the preparation of the hydrologic soil group (HSG)map, a digital text file of soil data was prepared to assign the soil data layers based on soil mappingunit. Spatial Analyst and XTools extensions of Arc View 3.3 were applied for map preparation. TheSoil Surveys from NRCS which provides a list of soil types and corresponding hydrologic soil groupswere used. The generated map contains individual polygons of the characterized hydrologic soil group.2.5. Generating CN mapTo create the CN map, the hydrologic soil group and land use maps were uploaded to the Arc Viewplatform. The Xtools extension of Arc View was used to generate the CN map. The hydrologic soilgroup field from the soil theme and the land use field from the land use map were selected forintersection. After intersection, a map with new polygons representing the merged soil hydrologicgroup and land use (soil-land map) was generated. The appropriate CN value for each polygon of the
579 Mahboubeh Ebrahimian, Lai Food See, Mohd Hasmadi Ismail and Ismail Abdul MalekSoil-Land map was assigned. The CN values for different land uses and hydrologic soil groups wereadopted from Technical Release 55, USDA-NRCS, 1986.2.6. Antecedent moisture condition (AMC)The calculated CN value for each polygon is for average conditions (i.e. Antecedent MoistureCondition Class II). The CN values for AMC II can be converted into CN values for AMC I and AMCIII by using the SCS standard tables (USDA-SCS, 1993). To determine which AMC Class is the mostappropriate relative to study area, the use of rainfall data is necessary. The 5-day rainfall prior to theevent date was determined to be used for converting the calculated CN value to AMC class II andAMC class III using based on the NRCS standard tables.2.7. Calculating runoff depth without incorporating the slope factorAfter generating CN map the next step was to calculate maximum potential retention (S). S value wascomputed for each polygon using equation (1). Runoff depth was ascertained for each rainfall event byusing equation (2). Arithmetic mean rainfall of available rain gauge stations in the watershed was usedfor estimation of runoff depth in the watershed for selected events. 25400 S= − 254 (1) CN ( P − 0.2S ) 2 Q= (2) P + 0.8Swhere Q is runoff depth (mm); P is rainfall (mm); S is potential maximum retention (mm); S is initialabstraction of rainfall by soil and vegetation (mm) and CN is Curve Number. At the next step, weighted runoff depth was estimated for the watershed by multiplying the areaof each polygon in runoff depth value and divided by total area of watershed (equation 3). In this study,a total of 35 daily rainfall events were employed in the NRCS-CN model to estimate runoff depth forthem. Q= ∑ QiAi (3) Awhere Qi is runoff depth for each polygon (mm); Ai is polygon area (ha) and A is watershed area (ha).2.8. Calculating slope-adjusted CNThe NRCS-CN method does not take into account the effect of slope on runoff yield. However, thereare few models which incorporate a slope factor to CN method to improve estimation of surface runoffdepth and volume (Huang et al., 2006). Fewer attempts have been made to include the slope factor intothe CN method. Those which had taken the slope factor into account were notably, Sharpley andWilliams (1990) and Huang et al. (2006). Equations (4) and (5) were used to adjust the CN valuesobtained from SCS-CN standard tables for the slope, respectively. Both methods assume very simplythat CN obtained from SCS standard tables correspond to a slope of 5%. CN2α= 1/3 (CN3-CN2) (1-2e-13.86α) +CN2 (4)where CN2α is value of CN2 for a given slope; CN2 and CN3 are the NRCS-CN for soil moisturecondition II (average) and III (wet), and α (m m-1) is the soil slope. CN2α= CN2 × K (5) Where: 322.79 + 15.63(α ) K= α + 323.52where K is a CN constant. Slope and CN maps were intersected to get slopes of each polygon. Since each polygon hasdifferent slopes, then calculating of weighted slope is need for each polygon. Weighted slope of a
Application of Natural Resources Conservation Service – Curve Number Method for RunoffEstimation with GIS in the Kardeh Watershed, Iran 580polygon was computed using formula (6). Weighted slope of polygon was applied in equation (5) tocompute slope adjusted CN values. The Huang et al. (2006) approach was used because of theimprovement made to incorporate the slope factor into the analysis. (6) Weighted slopewhere ai is area of slope (ha); si is slope (%); and A is polygon area (ha).2.9. Calculating slope-adjusted runoff depthThe same method as discussed above was employed to calculate slope adjusted runoff depth using ofslope-adjusted CN values for calculating S values. Accordingly, weighted runoff depth value wasestimated for the watershed for all rainfall events with corporation of slope factor.2.10. Determining runoff depth for observed dataDirect runoff volume was calculated by subtracting base flow and total runoff volume in WHAT(Web-based Hydrograph Analysis Tool) software (Engel et al., 2004). Runoff depth was calculated byequation (7) as following: 24 ∑ (Q − bf ) × t H= i −1 (7) Awhere H: is runoff depth (m); Q is runoff volume (m3/s); bf is base flow (m3/s); t is hourly time interval(3600) and A is watershed area (m2).2.11. Statistical analysisSPSS functions were used to perform statistical analysis on data. The Kolmogorov-Smirnov andShapiro-Wilk tests were used to check the normality of data set. The differences were consideredstatistically significant when p < 0.05. Percentage error was calculated to compare the differencebetween the estimated and observed runoff depth. Pair wise comparisons were done with the t-test tocompare observed and estimated runoff depth data. The spearman rank correlation analysis was used toinvestigate the relationship between estimated (as a dependent variable) and observed (as anindependent variable) runoff depths.3. Results and Discussion3.1. Hydrologic soil groupsAll hydrologic groups including A, B, C, and D were found in the Kardeh watershed (Figure3). Only 2percent of soil was placed in group A and about 40.6 and 31.7 percent of soil were placed in group Cand D, respectively (Table 2).
581 Mahboubeh Ebrahimian, Lai Food See, Mohd Hasmadi Ismail and Ismail Abdul MalekTable 2: Curve number of various land use and hydrologic soil groups in Kardeh watershed Land use/ Land cover Hydrologic soil group Area (ha) CN A 102.30 62 Dry farmland (rainfed farming) B 2204.90 71 C 4378.10 78 A 83.98 36 Forest B 1262.60 60 C 809.60 73 Thin forest D 398.70 79 B 24.40 55 Forest C 447.70 70 Fair forest D 44.16 77 A 4.70 35 Rangeland B 72.60 35 C 1208.54 47 Good condition D 1996.19 55 A 26.97 51 Rangeland B 1809.90 51 C 2628.26 63 Fair condition D 4808.70 70 A 215.36 67 Rangeland B 5649.30 67 C 7930.9 80 Poor condition D 6649.90 85 A 428.99 43 Orchards and B 510.30 65 irrigated farmland C 803.00 76 Settlement D 28.40 93 Rocks D 286.50 91 Total area 44814.95 -3.2. CN valuesThe CN value for each soil hydrologic group and corresponding land use class are presented in Table2. Hydrologic soil groups A and B leading to low CN value while the hydrologic group D leading tothe high CN value in the Kardeh watershed. Gandini and Usunoff (2004) observed that hydrologic soilgroup B leading to lower CN values in a humid temperate watershed of Argentina. In terms of land useand hydrologic soil group combination, the lowest CN value was found to be 35 and 36 in forest andrangeland with good condition and the highest CN value was found to be 93 in settlement areas.Gandini and Usunoff (2004) found the CN value of 92 for urban area and 45 for forest in goodcondition in Argentina. Table 2 indicates that rangelands with poor condition, settlements andmountainous areas without developed soil layer (rocks) are major contributors in runoff generation inthe Kardeh watershed. Nassaji and Mahdavi (2005) found that rangeland with poor and very poorconditions had CN values greater than 85 in three rangeland watersheds in semi-arid areas of northernIran. High CN value in poor rangelands can be explained by low vegetation density, high soilcompaction due to treading by grazing animals and low infiltration rate. The CN values map is displayed in Figure 4.The CN map can be viewed as a mosaic of CNvalues due to differences in land use. About 70 percent of Kardeh watershed has CN values between 60and 80, 4 percent less than 50 and 0.7 percent greater than 90. These values show that Kardehwatershed generates more runoff for a given rainfall in areas having greater CN values. Because byincreasing the value of CN in a specific area, the amount of runoff will be increased. Mellesse and Shih(2002) indicated that any changes in land use can alter CN values of the watershed and accordingly therunoff response of the watershed by increasing runoff volume. The study also reported that bydecreasing the area of croplands and rangelands within two decades the CN values greater than 90
Application of Natural Resources Conservation Service – Curve Number Method for RunoffEstimation with GIS in the Kardeh Watershed, Iran 582increased by 2.2 percent and the area of the watershed having runoff depth greater than 180 mmincreased by 2 percent. The relationship between the calculated CN from SCS Standard Tables and slope-adjusted CNwith land slope of the study area is shown in Figure 5. The figure suggesting that there is a directpositive relationship between slope and CN value. Higher CN values are expected in steep slope land.Slope-adjusted CN and standard CN ratio increases with slope. Slope-adjusted CN values are listed inTable 3. The highest CN value (93) was found in steep slope (32 %) of the study watershed while thelowest CN value (35) was found in slight slope (17 %) of the watershed. In watersheds where landslope is higher than 5%, CN values must be adjusted with slope (Huang et al, 2006).3.3. Runoff depthThe estimated runoff depths, slope-adjusted runoff depths and observed runoff for all selected rainfallevents are listed in Table 4. Comparison of columns 4 and 5 in Table 4 shows that there is no muchdifference between runoff depth before and after applying of slope factor. In the other words, afterusing this equation and incorporating slope factor in CN method, the CN values changed, but therunoff generation was not affected by new values considerably. This is largely due to the equation used(Huang et al, 2006). In this study, the equation used was developed over plot scale and the applicationis largely targeted for small sites (Huang et al, 2006). To date, there is very little information regardingmodification of NRCS-CN method for steep slopes at watershed scale. In fact, the equation used hereis the only available in the literature to modify runoff response with slope factor. The assessment of theeffect of slope on rainfall-runoff relationship in the NRCS-CN method is extended in this study toevaluate the effect on surface runoff generation at watershed scale. It is obvious that the curve numbervalues must be adjusted with slope degree to overcome to such problems in steep slopes. Figure 3: Hydrologic soil groups of Kardeh catchment.
583 Mahboubeh Ebrahimian, Lai Food See, Mohd Hasmadi Ismail and Ismail Abdul Malek Figure 4: Map of curve number values for Kardeh watershed.Figure 5: Relationship between the calculated CN from SCS standard tables and slope-adjusted CN ratio and land slope. 1.03 p<0.01 r = 0.75 1.02 Slope-adjusted/ tabulated CN ratio 1.02 1.01 1.01 1.00 1.00 0 10 20 30 40 50 60 Slope (%)
585 Mahboubeh Ebrahimian, Lai Food See, Mohd Hasmadi Ismail and Ismail Abdul MalekTable 4: Estimated runoff depths for rainfall events using NRCS-CN method Sum of prior 5- Estimated runoff Estimated slope-adjusted Observed Storm date Rainfall (mm) day rainfall (mm) depth (mm) runoff depth (mm) (mm) 14/5/1991 18.0 18.3 5.44 5.21 3.5 1/6/1992 17.0 0.2 5.71 5.47 8.2 11/7/1992 20.0 14.9 4.94 4.73 3.8 6/1/1993 26.1 29.7 6.56 7.55 5.6 8/3/1993 8.6 11.2 8.56 8.27 6.6 13/4/1993 22.9 4.4 4.30 4.12 11.0 7/5/1993 6.3 4.0 9.54 9.24 5.5 12/3/1994 11.0 22.2 0.92 1.21 4.6 14/6/1994 13.5 00 6.77 6.51 5.2 3/10/1994 5.9 2.8 9.72 9.42 5.6 1/5/1995 9.0 3.0 8.40 8.11 6.8 3/7/1995 19.0 9.6 5.18 4.96 2.4 4/2/1996 14.9 27.7 1.26 1.24 3.9 8/3/1996 17.7 44.9 2.77 3.40 5.0 14/3/1996 9.4 32.8 0.69 0.90 3.4 23/5/1996 6.6 9.8 9.41 9.11 7.2 27/5/1996 6.2 10.3 9.59 9.28 6.1 17/7/1996 23.5 00 4.18 4.01 3.3 6/5/1997 15.6 12.0 6.11 5.87 7.8 19/6/1997 24.1 7.9 4.07 3.90 7.3 1/8/1997 17.5 0.8 5.57 5.34 3.0 6/11/1997 8.0 14.3 2.34 2.19 1.5 9/2/1998 26.1 4.3 3.27 3.23 4.0 14/3/1998 6.1 1.2 7.73 7.47 5.9 26/3/1998 14.0 0.8 5.15 4.96 4.7 6/4/1998 25.1 2.6 3.34 3.29 5.3 27/4/1998 13.1 2.2 5.39 5.18 2.6 30/5/1998 7.8 4.5 7.07 6.82 4.3 22/7/1998 6.9 00 7.41 7.15 4.7 3/8/1998 5.3 0.5 8.07 7.80 7.5 14/8/1998 4.3 00 8.51 8.23 5.1 21/2/1999 27.1 3.8 3.20 3.18 4.4 28/4/2000 19.0 10.5 4.11 3.97 3.2 9/8/2000 8.1 1.1 6.96 6.71 5.2 18/8/2001 15.5 0.3 4.80 4.62 5.33.4. Comparison of estimated and observed runoff depthAt first step in the analysis, percentage error was used to compare the difference between the estimatedand observed runoff depth (Table 5). The maximum and minimum error between observed and estimatedrunoff depth were 115 and 7 percent, respectively. However, the maximum and minimum error betweenobserved and slope-adjusted runoff depth were 106 and 4 percent, respectively. The mean percent errorbetween observed and estimated runoff depth reduced from 46.26 % before adjusting for slope to 42.97% after adjusting for slope (Table 5). In India, Pandey et al. (2003) reported that the maximum andminimum error between observed and estimated runoff depth were 68.33 and 3.27 percent,respectively. Malekian et al. (2005) also reported an average percent error of 68.3 between observedand estimated runoff by the CN method for 25 storm events in semi-arid areas of north western Iran. Inthis study, about 9 % and 6 % of the estimated and slope-adjusted values were within ±10% of therecorded values, respectively. About 34 and 37 percent were within ±30 % of the observed runoff.About 43 and 37 percent of the estimated and slope-adjusted values were in error by more than ±50 %,respectively (Table 5). A percent error of less than 50 % was considered acceptable (Boughton and Chiew, 2007 andPandey et al., 2003). Statistical analysis indicated that percent error of estimated slope-adjusted runoff
Application of Natural Resources Conservation Service – Curve Number Method for RunoffEstimation with GIS in the Kardeh Watershed, Iran 586depth was significantly (P < 0.01) lower than the percent error of estimated runoff depth. This declinein percent error can be explained by the role of slope in runoff generation in mountainous watershed.One of the potential sources of error in runoff depth estimation is believed to be due to the rainfall andrecorded runoff data input. The quality of the input data is the main determinant of the quality of theresults in runoff estimation (Boughton and Chiew, 2007; Jacobs and Srinivasan, 2005). The presenceof various land use/ cover classes or condition in the watershed, and mountainous topography and largearea of the watershed may have played a part in the lack of acceptable runoff estimate results forselected storm events in this study. Field workers errors in recording rainfall and associated runoff dataprovide data with error to users which probably are another source of error. Pair wise comparison between the variable (observed vs. estimated runoff) means showed thatthere is no significant difference between means of estimated and observed data (P > 0.05). Therefore,estimated runoff depth by CN method was near to corresponding observed runoff depths (Table 6).Interestingly, Pandey et al. (2003) found that estimated direct runoff depth by the NRCS-CN methodwas significantly (P > 0.05) near to corresponding observed runoff depth in the Karso watershed, India.Similar results were reported by Pandey and Sahu (1999), Pandey et al. (2003) in India and Akhondi etal. (2001) in Iran.
587 Mahboubeh Ebrahimian, Lai Food See, Mohd Hasmadi Ismail and Ismail Abdul MalekTable 5: Details of percent error between estimated and observed runoff depth Percent Percent % of error error % of % of total total Study between % of between observe number of Acceptabi number Acceptabil storm slope observed estimated d storm lity of ity date adjusted and runoff and observed runoff events storm observed runoff events runoff 14/5/1991 7 4 0-10 5.7 Very high 1/6/1992 9 0-10 8.58 Very high 5 11/7/1992 9 12 6/1/1993 17 19 8/3/1993 18 19 13/4/1993 21 21 7/5/1993 23 24 12/3/1994 26 24 14/6/1994 27 24 10-30 37.3 high 10-30 34.30 high 3/10/1994 28 25 1/5/1995 29 25 3/7/1995 30 26 4/2/1996 30 26 8/3/1996 30 27 14/3/1996 30 29 23/5/1996 31 32 27/5/1996 33 33 17/7/1996 36 30-50 14.30 fair 34 6/5/1997 44 37 30-50 20 Fair 19/6/1997 45 46 1/8/1997 55 46 6/11/1997 56 48 9/2/1998 57 52 14/3/1998 57 52 26/3/1998 60 58 6/4/1998 64 61 27/4/1998 66 62 unacceptab 30/5/1998 68 > 50 43 68 le unacceptab 22/7/1998 73 68 > 50 37 le 3/8/1998 73 68 14/8/1998 80 73 21/2/1999 80 73 28/4/2000 85 78 9/8/2000 107 99 18/8/2001 115 106 Minimum = 7 Minimum = 4 Maximum = 115 100 Maximum = 106 100 Mean = 46.23 Mean = 42.97 The results showed that there is no provision to apply the NRCS-CN model in the Kardehwatershed for runoff estimation. It is noteworthy that the P values in Table 6 indicate that mean ofslope-adjusted estimated data (5.50) are nearer to mean of observed data (5.11) than estimated data(5.63) in terms of depth. Low P value means estimated and observed data are roughly far from eachother and vice versa. The comparison of observed runoff with estimated slope-adjusted runoff showed there is nosignificant difference between the means of observed and slope-adjusted estimated runoff (P > 0.05). Itshould be noted that for slope-adjusted runoff vs. observed runoff P value (0.27) was greater than Pvalue for estimated runoff vs. observed runoff (0.16) (Table 6). This means that when runoff depths
Application of Natural Resources Conservation Service – Curve Number Method for RunoffEstimation with GIS in the Kardeh Watershed, Iran 588were adjusted for slope, their means (5.50) were nearer to observed runoff depths (5.11). This indicatesthat slope is an important factor in runoff estimation. In steep slope watersheds, estimated runoff mustbe adjusted for slope since the estimations are affected more.Table 6: Means comparison of estimated and observed runoff for the Kardeh watershed Slope-adjusted P Estimated runoff Observed runoff Variable runoff depth Estimated vs. Slope-adjusted depth (mm) depth (mm) (mm) observed vs. observed Mean 5.63 5.50 5.11 0.16 0.27 SD 2.53 2.41 1.90 A fairly direct positive correlations was found between observed and estimated data (r = 0.55; P< 0.01) and slope adjusted vs. observed runoff data (r = 0.56; P < 0.01). In India, Nayak and Jaiswal(2003) found a good correlation (about 90 %) between estimated and observed data in all eight sub-basins with various areas (less than 100 km2) of the Bebas watershed, although correlation decreasedby increasing the area of the sub-basins. Akhondi (2001) pointed out that correlation coefficient (r)between observed and estimated runoff using the CN method decreased from 98 % to 17 % withincreasing watershed area and decreasing rainfall (from semi-humid to semi-arid) in four watershedswith various areas and climate in semi-arid and semi-humid areas of south western Iran. Furthermore,Malekian et al. (2005) reported a correlation coefficient of 73 % between observed and estimatedrunoff by the CN method in a semi-arid watershed of northwestern Iran. In the present study, a faircorrelation (about 55 %) between estimated and observed runoff depth could be attributable to the bigarea of the watershed. As discussed above, correlation is higher in small watersheds compared tobigger ones. Another reason behind this low correlation may be due to the use of a non-localized CNmethod in this study. The CN method parameters still have not been localized and modified based onIranian condition. This should serve as a caution to managers and researchers utilizing the CN methodfor hydrologic modelling in Iran (Khojini, 2001 and Malekian et al., 2005).4. ConclusionsThe following conclusions can be drawn from the study: • The incorporation of NRCS-CN model and GIS facilitates runoff estimation from watershed and can augment the accuracy of computed data. • Nevertheless there was no high significant correlation between estimated and observed runoff depth (r = 0.56; p < 0.01) in this study, but still there is no enough reason that curve number method should not be used for ungauged watersheds which do not have runoff records to produce runoff data for the purpose of management and conservation. • Although the results of this study failed to show the real effect of slope on runoff generation in the watershed due to the used equation (equation 5) at watershed scale, but the assessment of the effect of slope on rainfall-runoff relationship in NRCS-CN method should be extended further at watershed scale to get the effect on surface runoff generation. • In this study using of combined GIS and CN model to estimate runoff data in Kardeh watershed neither approved nor rejected completely. The results indicated that the combined GIS and CN method can be used in ungauged watershed with the same condition to Kardeh with about 60 percent accuracy only for management and conservation purposes not for computation of design floods.
589 Mahboubeh Ebrahimian, Lai Food See, Mohd Hasmadi Ismail and Ismail Abdul MalekAcknowledgementAuthors acknowledge University Putra Malaysia for providing research fellowship to make thisresearch possible. We are particularly grateful to Khorasan Razavi Regional Water Authority,Department of Watershed Management and Abkhiz Gostar Shargh Co. Ltd. for providing the data andmaps of the study watershed.References Akhondi, S., 2001. “An investigation of curve number model in flood estimation using Geographical information System (GIS)”, MSc thesis. Tarbiat Modares University, Tehran. Boughton, W., and F. Chiew, 2007. “Estimating runoff in ungauged catchments from rainfall, PET and the AWBM model”, Environmental Modelling & Software 22, pp. 476-487. Engel, B., J. Kyoung, Z.Tang, L. Theller, and S. Muthukrishnan, 2004. “WHAT (Web-based Hydrograph Analysis Tool)”, http://cobweb.ecn.purdue.edu/~what/ (accessed Feb. 2009). Gandini, M.L., and E.J. Usunoff, 2004. “SCS Curve Number Estimation Using Remote Sensing NDVI in A GIS Environment”, Journal of Environmental Hydrology 12, (Paper 16). Hossein, A.A., D.H. Pilgrim, G.W. Titmarsh, I. Cordery, 1989. “Assessment of U.S. Conservation Service method for estimating design floods. New direction for surface water modeling (Proceedings of the Balimore Symposium)”, IASH publication. No. 181, pp. 283- 291. Huang, M., G. Jacgues, Z. Wang, and G. Monique, 2006. “A modification to the soil conservation service curve number method for steep slopes in the Loess Plateau of China” Hydrological processes 20(3), pp. 579-589. Jacobs, J.H. and R. Srinivasan, 2005. “Effects of curve number modification on runoff estimation using wsr-88d rainfall data in Texas watersheds”, Journal of Soil and Water Conservation. 60 (5), pp. 274-278. Khorasan General Office of Natural Resources, 1993. “Watershed management plan for Kardeh dam watershed”, Report. 50p. Khorasan Razavi Jihad-e-Agriculture Organization, 1996. “Evaluation of land capability and soil of the Kardeh watershed”, Watershed Management Department, 150p. Khorasan Razavi General Office of Natural Resources, 2001. “Rangeland Vegetation Cover of Kardeh Watershed”, Report. 89 p. Khorasan Razavi Regional Water Authority, 2001. “Climatology of Kardeh Dam”, report. 180p. Khojini, A. 2001. “Investigation on the applicability of the SCS-CN method in runoff depth and peak discharge estimation in representative watersheds of Alborz Mountain chain”, Research and reconstruction 38, pp. 12-15. Malekian A., M. Saravi Mohseni and M. Mahdavi, 2005. “Applicability of the USDA-NRCS Curve Number method for runoff estimation”, Iranian Journal of Natural Resources 57(4), pp. 621-633. Melesse A.M., S.F. Shih, 2002. “Spatially distributed storm runoff depth estimation using Landsat Images and GIS”, Computers and Electronics in Agriculture 37, pp.173-183. Nassaji M. and M. Mahdavi, 2005. “The determination of peak-flood using different curve number methods (case study, Central Alborze area)”, Iranian Journal of Natural Resources 58(2), pp. 315-324. Nayak, T.R. and R.K. Jaiswal, 2003. “Rainfall-runoff modelling using satellite data and GIS for Bebas river in Madhya Pradesh”, IE (I) Journal 84, pp.47-50. Pandey, A. and A. K. Sahu, 2002. Generation of curve number using remote sensing and Geographic Information System”, http://www.GISdevelopment.net (accessed on Sep. 2007).
Application of Natural Resources Conservation Service – Curve Number Method for RunoffEstimation with GIS in the Kardeh Watershed, Iran 590 Pandey, A., V.M. Chowdary, B.C.Mal, and P.P. Dabral, 2003. “Estimation of runoff for agricultural watershed using SCS curve number and Geographic Information System”, Htpp://www.GISdevelopment.net (accessed on Sep. 2007). Sharpley, A. N. and J.R. Williams, 1990. “EPIC- Erosion/Productivity Impact Calculator: 1. Model determination”, US department of Agriculture. Technical Bulletin, No. 1768. USDA/NRC. 1986. “Urban Hydrology for Small Watersheds TR-55”, Technical Release 55. Washington DC. USDA-SCS. 1993. “Storm rainfall depth. In: National Engineering Handbook Series”, Part 630, Chapter 4, Washington, D.C. Zhan, X. and M. L. Huang, 2004. “ArcCN-Runoff: An ArcGIS tool for generating curve number and runoff maps”, Environmental Modelling & Software 19, pp.875–879.