ing, and has been incorporated into various computer models worldwide (Woodward et al., 2002; Hawkins et al., 2002). Although this is an accepted method for runoff estima- tion, studies have indicated that it should be Effects of curve number modification on evaluated and adapted to regional agro- climatic conditions. Hawkins (1998) states that curve number runoff estimation using WSR-88D rainfall tables should be used as guidelines and that actual curve numbers and their empirical data in Texas watersheds relationships should be determined based on local and regional data. This is supported by J.H. Jacobs and R. Srinivasan Van Mullem et al. (2002). They state that the direct runoff calculated by the curve ABSTRACT: The purpose of this study was to evaluate variations of the Natural Resources number method is more sensitive to the Conservation Service (NRCS) curve number method for estimating near real-time runoff for curve number variable than rainfall inputs. naturalized flow, using high resolution radar rainfall data in Texas watersheds. This study was This would suggest an increased need for undertaken in an attempt to provide more accurate runoff estimates to watershed and water field verification of land cover type and con- resource managers for planning purposes. Stage III Weather Surveillance Radar 1988 Doppler dition before curve number assignment. The accuracy of hydrologic models (WSR-88D) precipitation data, obtained from the West Gulf River Forecast Center, was used as depends on the accuracy of input data and, in the precipitation input for runoff estimates in this study. The study areas consisted of dominant the case of the NRCS curve number or homogenous land use and were characterized by naturalized flow. Findings indicate that the method, the variable inputs. The purpose of use of a dry antecedent soil moisture condition curve number value and a reduced initial this study was to evaluate several variations of abstraction coefficient (Ia) in the NRCS curve number equation produced the most statistically the NRCS curve number method for esti- significant comparison between observed and estimated runoff in nine out of ten watersheds. mating near real-time runoff for naturalized The combined comparison for all events in these nine watersheds produced a coefficient of flow, using WSR-88D rainfall data for water- efficiency (COE) of 0.70, with a slope of 0.78 and an r2 of 0.77. Overall, the results of this sheds in various agro-climatic regions of research suggest that, although further improvements can be made for improved runoff Texas in an attempt to represent the spatial estimation, the use of modified inputs to the NRCS curve number equation in conjunction with variability of rainfall. WSR-88D radar rainfall data could be useful in producing runoff estimates for Texas in real-time. Methods and Materials Study area selection and description. Te n Keywords: Hydrologic modeling watersheds of varying size (355 to 2,940 km2), in four river basins, throughout differ- ent agro-climatic regions of Texas were used Water availability has become a major in arid and semi-arid regions, such as west in this study to account for a variety of issue in Texas in recent years. Adding to Texas, where most rainfall occurs in short, hydrologic conditions throughout the state this issue is the expected doubling of the pop- heavy, localized thunderstorms. Dense net- (Figure 1). Watersheds were chosen based on ulation within the next 50 years, mainly in works necessary to provide such data are gen- dominant land use, soil hydrologic group, and areas presently without abundant water sup- erally available only for experimental or streamgauge location (Table 1). Land cover plies. To combat problems that Texas will research watersheds. Also, few rain gauge data was obtained from the 1992 U.S. face in the future, there has been a move networks are currently able to provide real- Geological Survey (USGS) National Land toward active planning and management of time data. The use of weather radar systems Cover Dataset at a 1:24,000 scale (30 m water resources. Real-time weather data could help alleviate these problems. One resolution). Only watersheds with homoge- processing and hydrologic modeling can pro- such system is the Weather Surveillance nous/dominant cover or similar curve num- vide information for this planning in addition Radar 1998 Doppler (WSR-88D) of the ber values were used. Soils data were derived to flood and drought mitigation, reservoir National Weather Service (NWS). This data from the U.S. Department of Agriculture operation, and watershed and water resource could be used in conjunction with runoff Natural Resources Conservation Service management practices (Texas Water Develop- models, such as the Natural Resources (USDA-NRCS) State Soil Geographic ment Board, 2000). However, in order to Conservation Service (NRCS) curve number (STATSGO) database, at a 1:250,000 scale provide this information to managers, it is method to deliver data to managers in a near (250 m resolution). Streamflow data was necessary to first obtain reliable and readily- real-time fashion. available weather data. In the 1950’s, the curve number method Jennifer H. Jacobs is a research associate and Rain gauge networks are generally sparse was developed to estimate runoff in Raghavan Srinivasan is the professor and director and insufficient to capture the spatial variabil- ungauged watersheds (SCS, 1972). This at the Spatial Sciences Laboratory at Texas A&M ity of rainfall across large watersheds, especially method is widely used for watershed model- University in College Station, Texas. Reprinted from the Journal of Soil and Water Conservation 274 JOURNAL OF SOIL AND WATER CONSERVATION S|O 2005 Volume 60, Number 5 Copyright © 2005 Soil and Water Conservation Society Figure 1 Study area locations. Red-1 (3) For the actual runoff calculation, initial abstractions (Ia) are generally approximated as Red-2 0.2 retention parameter, and the basic equa- Trinity-1 tion becomes (Equation 4): (4) where, LCR-1 Qsurf = surface runoff in mm, Trinity-2 Trinity-3 Rday = rainfall depth for the day,also in mm. LCR-2 Runoff will occur only when Rday greater LCR-3 than Ia (Neitsch et al., 2001). However, Ponce and Hawkins (1996) suggest that 0.2 SA-1 SA-2 retention parameter may not be the most N W E appropriate number for initial abstractions, 0 50 100 200 S and that it should be interpreted as a regional Kilometers parameter. Stage III WSR-88D data was obtained downloaded from the USGS website for each defined as (Equation 1): through a Memorandum of Agreement with watershed outlet. the West Gulf River Forecast Center Total streamflow is composed of baseflow (WGRFC) of the National Weather Service (lateral flow and shallow ground water dis- (1) (NWS). Data for the 1999 - 2001 time period charge to streams) and surface runoff. To was used as the rainfall input in this study, compare measured flow and estimated runoff based on findings by Jayakrishnan (2004) (NRCS curve number method provides only Curve number varies based on one of citing improved data quality and accuracy in direct runoff after a rainfall event), it was nec- three antecedent soil moisture conditions, recent years. This, in addition to the fact that essary to determine the portion of streamflow curve number I-dry (wilting point), curve the WSR-88D data is complete and available that could be attributed to surface runoff. number II-average, and curve number III-wet daily, makes it a more useful dataset for this Therefore, flow data was processed through a (field capacity) (Neitsch et al., 2001). Runoff type of modeling research. More informa- filter program. The filter program developed estimates increase with increasing antecedent tion concerning weather radar products and by Arnold et al. (1995) was used in this study, soil moisture condition, and with increasing processing algorithms can be found in Crum and is comparable to other automated separa- curve number. and Alberty (1993), Klazura and Imy (1993), tion techniques, with 74 percent efficiency Curve number II was assigned based on Smith et al. (1996), and Fulton et al. (1998). when compared to manual separation the dominant land use from National Land Comparing runoff results. For this analysis, (Arnold and Allen, 1999). Also, because the Cover Dataset and soil hydrologic group from curve number I and curve number II were runoff algorithms used in this study do not STATSGO according to the Soil Conserva- first used as curve number variables in the account for reservoirs or other diversions, tion Service’s (SCS) Texas Engineering runoff equation with an initial abstraction only sites with natural, or unregulated, flow Technical Note No. 210-18-TX5 (1990). ratio of 0.2, 0.1, and 0.05 to determine the were used. This allowed for a direct compar- Curve number I and curve number III are most appropriate constant for initial abstrac- ison of runoff estimates to measured stream- calculated from curve number II and are tions in the selected study sites. WSR-88D flow data. defined by Equations (2) and (3) respectively data was used as the rainfall input. Results of Curve number calculations. Daily runoff (Neitsch et al., 2001). A geographic informa- each alternative (a combination of curve calculations were generated using the NRCS tion system (GIS) layer in grid format was number variable and initial abstraction coeffi- curve number method. This calculation is created for each watershed based on curve cient) were then evaluated based on observed based on the retention parameter, S, initial number II values at a 4 km resolution, from runoff to determine which produced the abstractions, Ia (surface storage, interception, which curve number I was calculated. most statistically significant results (Hadley, and infiltration prior to runoff), and daily 2003). rainfall, Rday, (all in mm H20).