Tropical Agricultural Research Vol. IS : 307- 31S (2003) Remote Sensing and GIS based Methodology for Curve Number Estimation in Rainfall - Runoff Modelling I.B.J. Bandara, R. P. De Silva1 and R. K. Singh2 Dept. of Agricultural Engineering Faculty of Agriculture, University of Peradeniya, Sri Lanka ABSTRACT. This study was conducted using satellite data, image processing techniques and GIS functionality to estimate the flow regime of a catchment using USDA Soil Conservation Service (SCS) Runoff Curve Number (CN) method. The selected catchment for the study was the Bainkhala watershed, Doon valley, Uttaranchal state, India. An IRS IC LISS-III image was used to generate all the thematic layers of spatial data. The CN under different land use scenarios were adjusted by trial and error in the hydrological modeling process. Subsequently, these estimated curve numbers were used for the estimation of flow statistics. It was found that the different land use classes and the corresponding curve numbers were less sensitive to changes in the calculated runoff and the interactions between land use and soil basically determine the conversion efficiency of runoff generation from storm rainfall. Remotely sensed data, image processing techniques and basic GIS functionality are an efficient collection to estimate runoff curve numbers in the runoff estimation process. INTRODUCTION A catchment or watershed is defined as the spatial extent which contributes water to a selected location of the stream network. A watershed could encompass a few hectares as in the case of a small pond or an intermittent stream to thousands of hectares in the case of perennial rivers. Within each watershed, sub basins could be identified based on the elevation of each point on the stream network. Each watershed or sub watershed is an independent hydrological unit and hence it is easily manageable. Runoff is the output of a watershed in terms of water yield of the hydrological unity. The measurement of runoff is a very tedious process and therefore, estimation methods are commonly employed. The most common runoff estimation methods include rational formula. Cooks method, Hydrograph method, Empirical formula method and USDA SCS Runoff Curve Number method. The SCS CN method is a simple, widely used and efficient method for determining the approximate amount of runoff resulting from a single storm event over a small watershed. Although this method has been designed for a single storm event, it can also be scaled to find average annual flow generations. The data and information- requirements for this method are very low and require rainfall depths and curve numbers. The curve number can be estimated from the hydrologic soil group of the area, land use, treatments used on the land and hydrologic condition. The empirical equation requires rainfall depth and a coefficient for the watershed (curve number) as inputs (Juracek, 2002). Accurately 1 Department of Agricultural Engineering, Faculty of Agriculture, University of Peradeniya, Sri Lanka. 1 Scientist/Engineer-"SD". Water Resources Division, I IRS, India. Bandara el al. estimated runoff rates and runoff volumes can.be used for water supply forecasting, flood prediction and warning, hydropower production estimation, navigation management and many other water resources applications. The main objective of the study is to explore the possibilities of employing remote sensing methodologies and GIS functionality to estimate runoff using USDA SCS runoff curve number method. The specific objectives include (a) creation of spatial database for hydrological modeling (b) establishment of rainfall- runoff relationship based on observed runoff data and (c) creation of curve numbers for the different hydrological unit and its sensitivity analysis. DATA AND METHODOLOGY Description of the study area The study area covers Bainkhala watershed located in the western Doon valley, Uttaranchal state, India. It includes an area of approximately 16 km2 and extends from 30° 21' 04.71" N to 30° 24' 53.79" N latitude and 77° 52' 05.14" E to 77° 57' 05.16 E longitude. The watershed is bounded in the north by lesser Himalayan mountains, in the south ,by watershed of Asan river, in west by watershed of Suarna river and east by watershed of Barawala river. The Bainkhala River is basically a dry riverbed and flow is only found seasonally and hence could be categorized as non- perennial. The drainage pattern is sub-dendritic in upper part of watershed and braided in the lower part. The outlet of watershed joins the Asan River in the south. Rainy season is from mid June to September and annual rainfall is in the range of 1750 - 2000 mm (Bomjan, 1999). The dominant land uses in the study area are forest and agriculture. Forest areas are homogeneous Sal forest, which are spread in the eastern side of the watershed. The major agricultural crops are rice, wheaf 'a"hd sugarcane. The agricultural fields are spread throughout the western side. The other" prominent land use classes include Open forest, Scrubland, Orchard (Mango), Built-up lands and Riverbed. Open forest also mainly consists of Sal trees but the density is low and hence cannot be categorized into regular forest category. The prominent plant species of Scrubland is Lantana. There are three major soil texture classes identified as loamy skeletal, coarse loamy, fine loamy in the study area. Mostly A, B and C type of hydrological soil groups have been formed (Bomjan, 1999). Data used The data used in this study can be categorized into three (3) groups. Satellite data which include PAN and LISS-III merged data hard copy (path - 96, raw - 50) acquired on December 27. 2001 and LISS-III image (path - 96, raw - 50) acquired on February 1|,.,,1,999. The topographic data were obtained from Survey of India Toposheet N0..53F/15 on 1:50,000 scale (First edition, 1970) and from soil texture map prepared by Central Soil and Water Conservation Research and Training Institute, Dehradun. Further, 45 GPS points of the study area with ground truth were also collected. The. .meteorological data include daily Rainfall data at the Bainkhala meteorological station from July to September 1990 and runoff discharge data of the Bainkhala watershed at the gauging site from July to September, 1990. 308 http://can.be Remote Sensing and GIS Based Methodology In the processing of these satellite data, ERDAS Imagine 8.3 was used for geo referencing images and topographic map, and ILWIS 3.0 was used for some image processing and GIS analysis. The generic structure of the methodology adopted for this study is indicated in Figure 1. Satellite data were used to derive the land use status of the watershed through the unsupervised classification process. The accuracy of the classification was verified with GPS ground data. The derived land use map is shown in Figure 2. Contour information available on the toposheets and the demarcated boundary were transferred to obtain the Digital Elevation Model (DEM) and Slope aspect information. Attributes obtained from soil texture map were used along with the land use and slope data to determine the spatial distribution of hydrologic soil groups. Rainfall data Runoff data Mathematical formulae Satellite data Toposheet Unsupervised classification S values CN values I Boundary & Contours Land use map DEM/Aspect maps Cross South/North aspect Soil texture map Cross I Attribute data HSG map LUT & Soil/Soil CN values w e i g h t e a : c N ; u : ... . I J Calculated Runoff Fig. 1. Flow chart for the methodology. 309 Bandara et dl Landuse map of Bainkhala watershed ( 1 Airr ulttitp H i Ruill up I and I 1 f)p<«n fnr*>«f I I Orchard r l Shrub Fig. 2. Land use map of Bainkhala watershed. The Storage capacity (S) values for the watershed were calculated using measured rainfall and runoff data. I he Curve Number viilues for each rainfall and rtinolT event were calculated using these S values and the average value of the curve numbers was taken ns the Weighted Curve Number ns show in the following formulae. hence, Q (I' la)' ( / ' - la) + S (/' - l a ) 2 0(1' la) it — 7—• " - ('Nil = Q 25400 .V - 254 (la - ( L I S ) where. Q = Accumulated storm runofl (mm) P " Accumulated slorm rainfall (mm) S = Potential maximum retention of water hy the soil la ~ Initial abstraction before ponding .tio Remote Sensing and GIS Based Methodology RESULTS AND DISCUSSION The curve number was taken as CNII and then CNI & CNIII were calculated using the CNII value. The Antecedent Moisture Conditions classes were determined by considering the status. The details are given in Table 1. Table 1. Antecedent Moisture Classes Classification for Bainkhala watershed. . . AMC Moisture (mm) Condition class 1 < 35.6 Dry soil but not the wilting point . 11 35.6- 53.3 Average conditions III > 53.3 Saturated soils; heavy rainfall or light rain Based on the methodology of the USDA SCS CN method, the weighted curve numbers were calculated and are presented in Table 2. Table 2. Calculated weighted CN v: dues. RF (mm) RO (mm)-Observed SI S2 CN(S1) CN(S2) 217.8 53.83 1627 244 14 51 16.6 12.37 203 4 56 99 53 12.13 385 62 40 80 60 11.06 406 80 39 76 25.2 11.8 244 15 51 94 34.2 9.65 268 35 49 88 7.2 9.45 123 -1 67 101 11.2 13.74 183 -2 58 101 20.8 ' 14.37 244 6 51 98 2.6 10.76 103 -2 71 101 46.8 18.45 420 35 38 88 9 4.69 92 5 73 98 10.9 13.4 179 -2 59 101 105.6 52.99 1058 58 19 81 3.6 7.45 84 -2 75 101 81 18.41 587 96 30 73 21.2 22.49 317 -1 44 100 CNII = 49 CNI = 29 CNIII = 69 For each rainfall data two values for S (SI and S2) were obtained. S cannot be a minus value and CN cannot be above 100. Therefore, only SI values were taken for calculation. Further, in order to get the calculated weighted curve number (CNII), the curve numbers for different land use classes were adjusted. This was a trial & error method and the results are given in Table 3. Accordingly, the estimated CN map was generated as shown in Figure 3. 311 Bandara el al. Table 3. Adjusted CN values for different areas. Land .use Soil types HSG CN NPix Area (m2) S Area*CN Dense forest Coarse loamy B 54 II 7436 216.37 401544 Dense forest Loamy skeletal A 29 10054 6796504 621.86 197098616 Open forest Fine loamy- C 64 6 4056 142.88 259584 Open forest Coarse loamy B 59 8 5408 176.51 319072 Open forest Loamy skeletal A 49 2533 1712308 264.37 83903092 Agriculture Fine loamy C 84 87 58812 48.38 4940208 Agriculture Coarse loamy B 80 127 85852 63.50 6868160 Agriculture Loamy skeletal A 71 7628 5156528 103.75 366113488 Scrubland Fine loamy C 54 4 2704 216.37 146016 Scrubland Coarse loamy B 46 894 604344 298.17 27799824 Scrubland Loamy skeletal A 32 353 238628 539.75 .7636096 Riverbed Coarse loamy B 96 454 306904 10.58 29462784 Riverbed Loamy skeletal A 89 173 116948 31.39 10408372 Built-up Land Coarse loamy B 71 682 461032 103.75 32733272 Built-up Land Loamy skeletal A 56 80 54080 199.57 3028480 Orchard Coarse loamy B 59 1 676 176.51 39884 Orchard Loamy skeletal A 54 331 223756 216.37 12082824 Total 15835976 783241316 W. CN(CNll) 49 CNI = 29 CNIII = 69 The study was extended to find the relationship between observed runoff values and calculated runoff values. The results are shown in Figure 4. Based on the composite CN values determined for the watershed, rainfall-runoff modeling was attempted as shown in Figure 5. According to the modeling.results as shown in Figure 5, there is no perfect agreement with the measured and predicted runoff. It indicates that further revisions of composite runoff curve number values are necessary to obtain better modeling results. The study needs to be extended to several more seasons to obtain more realistic composite curve numbers.'4' 312 Ronnie Nomina nn skeletal and MSG A group contributes for comparatively low CN values of the watershed thus indicating that these areas have low runoff potential. Similarly. Agriculture and Riverbed land use categories along with fine loamy, coarse 314 Remote Sensing and GIS Based Methodology loamy or loamy skeletal soils and HSG groups A, B or C generate higher curve numbers indicating high runoff potential areas. For this study, a limited number of available runoff event data (17 observations) was used. It is useful to extend the runoff records to cover a large number of events to improve the results. Soil texture map used for this study was at a small scale. It is required to study Soil texture in detail on a large scale to improve the results. From the results of the study, it can be concluded that remote sensing and GIS techniques are efficient and are comparatively easy methods to estimate runoff curve numbers when compared with the conventional methods. However, it is noted that different land use classes and curve numbers are less sensitive to changes in the calculated runoff. In estimation of curve numbers, the spatial distribution of dynamic variables need to be taken into consideration and hence, a combination of remote sensing data sources and GIS analytical capabilities could provide the required temporal and spatial resolution for the data and their analysis. , REFERENCES Bomjan,. S. (1999). Remote sensing and GIS approach for rainfall - runoff modeling. . . University of Delhi. r. • . . • • . Chaurey, R. (2000). Rainfall runoff relationship in ,darer watershed.. Department, of. .• Geology. Mahathma Gandhi Chitrakoot Gramodaya Vishwavidyalaya.. Chow, V.T., David, R. and Maidment, Larry W. Mays. (1988). Applied Hydrology. McGraw-Hill series in water resources and environmental engineering. Impacts of Land Use Change on Water Resources. http//.ecn.purdue.edu/runoff/lndex Juracek. K. E. (2002). Estimation and comparison of Potential Runoff - Contributing areas in Kensas :using Topographic Soil and Land-Use Information., U. S. Geological Survey. , , . . . . . . . http://ks.watef.usgs.gov/Kensas/pubs/reports/,wrir.0p;i4177.htm l#HDR.l •••< • •• u: • : .. Mapsoflndia.www.mapsolindia.com . .}• , . . Ragunath, H.M. (2000). Hydrology - Principles Analysis and Design Runoff. New Age. International (P) Limited Publishers. , • • • Ward, A.D. and William J. Elliot. (1995). Environmental .Hydrology, Surface Runoff and Subsurface drainage. Lewis Publishers.;/. http://ks.watef.usgs.gov/Kensas/pubs/reports/,wrir.0p;i4177.htm http://Mapsoflndia.www.mapsolindia.com