Coupling HEC-HMS with Atmospheric Models for Prediction of Watershed Runoff M. L. Anderson, M. ASCE1; Z.-Q. Chen, M. ASCE2; M. L. Kavvas, M. ASCE3; and Arlen Feldman, M. ASCE4 Abstract: The operation of reservoirs in the Sierra Nevada mountains of California for flood control relies on forecasts of reservoir inflows. In the past, accurate forecasts of the reservoir inflows resulting from watershed runoff have been made, but only after the water has entered the main channel. During flooding events, this limits the amount of time available for the implementation of emergency management procedures. Translating precipitation forecasts into runoff forecasts can greatly improve the runoff-forecast lead time. The operational National Center for Environmental Prediction Eta model provides 48-h-ahead forecasts of precipitation in 6-h intervals in a 40⫻40 km gridded form. In this study, the mesoscale model, MM5, is used to transfer the Eta forecast data down to the appropriate space and time scales required to link the Eta model precipitation forecast results to the watershed model, HEC-HMS, for runoff prediction. An initial diagnostic study of this procedure has been performed on the Calaveras River watershed in Northern California. Initial results indicate that: 共1兲 model parameterization choice in MM5 is necessary to refine the precipitation forecasts; 共2兲 the method shows promise for generating 48-h-ahead forecasts of reservoir inflows; and 共3兲 calibration of the HEC-HMS model with distributed precipitation is necessary for this methodology. This paper presents the study results along with a discussion of the methodology. DOI: 10.1061/共ASCE兲1084-0699共2002兲7:4共312兲 CE Database keywords: Runoff forecasting; Watersheds; California. Introduction The operation of reservoirs in the Sierra Nevada mountains of California for flood control rely on forecasts of reservoir inflows. In the past, accurate forecasts of the reservoir inflows resulting from watershed runoff have been made, but only after the water had entered the main channel. During flooding events, this limits the amount of time available for the implementation of emergency management procedures. An example of this limitation and its consequences occurred during the December 1996/January 1997 flooding events over Northern California 共National Weather Service 1997兲. This study examines the possibility of increasing runoff-forecast lead time through the use of precipitation forecasts. Recent efforts in river forecasting have focused on quantifying rainfall amounts from radar images 共Charley 1986; James et al. 1993; Coontz 1994; Mimikou 1996兲. Research has also progressed in the continued development of models capable of predicting the spatial and temporal evolution of the flood wave as it 1 Researcher, Civil and Environmental Engineering, Univ. of California at Davis, Davis, CA 95616. 2 Researcher, Civil and Environmental Engineering, Univ. of California at Davis, Davis, CA 95616. 3 Professor, Civil and Environmental Engineering, Univ. of California at Davis, Davis, CA 95616. 4 Head, Research Division, United States Army Corps of Engineers, Hydrologic Engineering Center, Davis, CA 95616. Note. Discussion open until December 1, 2002. Separate discussions must be submitted for individual papers. To extend the closing date by one month, a written request must be filed with the ASCE Managing Editor. The manuscript for this paper was submitted for review and possible publication on December 3, 1999; approved on October 16, 2001. This paper is part of the Journal of Hydrologic Engineering, Vol. 7, No. 4, July 1, 2002. ©ASCE, ISSN 1084-0699/2002/4-312–318/$8.00⫹$.50 per page. moves down the channel 共Yapo et al. 1993; Franchini and Lamberti 1994; Lamberti and Pilati 1996兲. However, these reported studies are limited to the time frame of radar images and water that is already in the main channel. The only way to gain additional lead time in runoff forecasting is to gain precipitation information ahead of its occurrence. One way in which this can be accomplished is by translating precipitation forecasts into runoff forecasts. The National Center for Environmental Prediction 共NCEP兲 Eta model 共Staudenmaier 1996a,b兲 provides 48-h-ahead forecasting of precipitation in 6-h intervals in a 40⫻40 km2 gridded form over the entire United States. Through the use of a mesoscale model, MM5, and a rainfall-runoff model, HEC-HMS, this information can be translated into runoff forecasts with a 48-h lead time. Accurate runoff forecasts of this nature would greatly improve the lead time necessary for emergency management procedures such as evacuations to be carried out. A perfect forecast of runoff peak, volume, and timing would be the ideal scenario for such a methodology. However, effective modifications to emergency management procedures can be made with a less than perfect forecast. The implementation of a series of emergency management measures may be made as the forecast inflows increase. Because there is a range of flows that apply to a given set of measures, the runoff forecast should be accurate enough to fall into the correct flow range. That is, the predicted inflow and the observed inflow should trigger the same set of emergency management measures. As a means of evaluating this approach, a feasibility study has been completed on the Calaveras River watershed in Northern California. Eta model forecasts were obtained to provide input and boundary conditions for the mesoscale model, MM5. The MM5 model was then run to provide a refined precipitation forecast with spatial and temporal scales suitable for use in the United States Army Corps of Engineers 共USACE兲 watershed model HEC-HMS. HEC-HMS was run with the precipitation forecast 312 / JOURNAL OF HYDROLOGIC ENGINEERING / JULY/AUGUST 2002 Downloaded 28 Feb 2011 to 152.3.102.242. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org Table 1. Grid Points of Corners of AWIPS Grid #212 (I, J) 共1, 1兲 共1, 129兲 共185, 129兲 共185, 1兲 Fig. 1. Schematic diagram of process for obtaining precipitation forecasts. Runoff forecasts from HEC-HMS. data in order to provide a 48-h forecast of runoff entering a reservoir on the Calaveras River. The results of this study are presented in this paper, which starts with an overview of the methodology used to translate the Eta model precipitation forecasts into HEC-HMS runoff forecasts. After describing the computer models MM5 and HEC-HMS that are used in the study, a description of the Calaveras River watershed used in the case study is presented, including its representation in HEC-HMS. Results of the study are then presented and discussed. The paper concludes with an assessment of the methodology including future directions for its further development. Methodology and Model Descriptions A schematic of the methodology for translating precipitation forecasts into runoff forecasts is shown in Fig. 1. Eta-model products are first downloaded from NCEP’s Web site. This information provides the input and boundary conditions for the mesoscale model, MM5, which can then be considered to be nested into the Eta model. The mesoscale model, MM5, is run to produce 48 one-hour forecasts of precipitation depths on a grid covering the Calaveras River basin that will be used by the watershed model, HEC-HMS. In the following subsections, the Eta model forecast products, MM5, and HEC-HMS models are described, including the input that is necessary for this simulation methodology. Eta Model Forecast Products A series of operational Eta models for atmospheric numerical forecasts have been developed for use in the U.S. 共Burks and Staudenmaier 1996; Janish and Weiss 1996; Schneider et al. 1996兲. The gridded Eta model results can be found over the AWIPS 共Advanced Weather Interactive Processing System兲 grids Latitude/longitude 12.190 N/133.459 W 54.536 N/152.856 W 57.290 N/49.385 W 14.335 N/65.091 W in GRIB format. GRIB 共Gridded Binary兲 is a general purpose, bit-oriented data exchange format that is an efficient vehicle for transmitting large volumes of gridded data to automated centers over high-speed telecommunication lines using modern protocols. A GRIB decoder is required to read and to process the forecast data in their raw formats. NCEP releases two forecasts at 00Z and 12Z UTC 共Coordinated Universal Time兲 each day. The letter Z means that the time is the local time at the zero degree longitude 共Greenwich meridian兲. For hydrological application, it may be necessary to convert the UTC time to a local time. Each forecast has a lead time of 48 h with time intervals of 6, 3, or 1 h. The 6-h interval data is available to the public from the NCEP Data Repository Site at ftp.ncep.noaa.gov. The operational Eta model at the time of the study has a 32 km horizontal resolution and 45 vertical layers, and it runs over a domain that encompasses nearly all of North and Central America, including surrounding oceans and Alaska and Hawaii. Initial conditions are provided by the Eta Data Assimilation System 共EDAS兲, which runs on a 3-h forecast/analysis/update cycle for 12 h prior to the start time of a model run. Boundary conditions are provided by the previous cycle’s Aviation model 共AVN兲 run of the NCEP Global Spectral Model. NCEP uses an Eta model postprocessor to generate the outputs over several AWIPS 共Advanced Weather Interactive Processing System兲 grids that are more useful to meteorologists and hydrologists than the original Eta model outputs. The gridded Eta model output are further grouped into 3D output files and surface output files. AWIPS grid #212 was used for this study. This grid is defined on a Lambert Conformal projection with a nominal horizontal grid resolution of 40 km and grid dimensions of 185⫻129. The latitude/longitude locations of the corner points of this grid are given in Table 1. This domain covers most of North America and the nearby oceans, including the 48 contiguous United States, the southern half of Canada, and most of Mexico. Mesoscale Model MM5 The MM5 modeling system is the fifth generation mesoscale model developed by the National Center for Atmospheric Research 共NCAR兲 and the Pennsylvania State Univ. It is a globally relocatable model that can be run under either hydrostatic or nonhydrostatic dynamic frameworks. The nonhydrostatic Table 2. MM5 Parameterizations used in Forecast Study MM5 parameterization Atmospheric radiation Boundary layer Soil model Precipitation—nonconvective Precipitation—convective Scheme chosen Dudhia Blackadar Multilayer Explicit three-phase Grell JOURNAL OF HYDROLOGIC ENGINEERING / JULY/AUGUST 2002 / 313 Downloaded 28 Feb 2011 to 152.3.102.242. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org Table 3. Rain Gauges used in Calibration of Calaveras Watershed HEC-HMS Model Rain Gauge Esparanza Railroad Flat Sheep Ranch New Hogan Reservoir Robidart Ranch Latitude (°) Longitude (°) 38.242 38.314 38.210 38.155 38.137 ⫺120.497 ⫺120.543 ⫺120.462 ⫺120.814 ⫺121.030 the NCEP Eta model forecast on AWIPS grid #212, which has a spatial resolution of approximately 40 km. The AWIPS data set provides three-dimensional forecasts of temperatures, wind vectors, relative humidity, geopotential heights, and surface pressures every 6 h for a total of 48 h. Nested within the outer grid is a 34⫻34 cell inner grid with a resolution of 4 km. The inner grid spans the Calaveras River basin and its immediate surrounding areas 共Fig. 2兲 and uses relaxation boundary conditions. The computational time interval is 20 s and the time interval of the MM5 precipitation output is set to one hour to coincide with the data requirements of HEC-HMS. The parameterizations used in the MM5 simulation of the Calaveras basin are shown in Table 2. For this initial study, an evaluation of the impact of using different parameterizations within MM5 was not performed. A complete description of the parameterizations available in the MM5 modeling system can be found in Grell et al. 共1994兲. Fig. 2. Map of California showing location of Calaveras River and MM5 model outer and inner grid boundaries framework, used in this study, allows the model to be used at a few-kilometer scale 共Dudhia 1993兲. The state variables of the MM5 modeling system with the nonhydrostatic framework include pressure, temperature, density, and wind velocities. Parameterized processes include advection, diffusion, radiation, boundary-layer processes, surface-layer processes, cumulus convection, and routines for all three phases of water in the atmosphere. Several options exist for the parameterization of moist convection and boundary layer processes for the simulation of atmospheric phenomena at different scales and different characteristics 共Kavvas and Chen 1998兲. The MM5 model for the Calaveras River basin is a 32-layer model with two nested grids identified as the outer grid and inner grid. The 31⫻31 node outer grid has a spatial resolution of 12 km and spans central California, part of Nevada near Lake Tahoe, and part of the Pacific Ocean by San Francisco and Half Moon Bays 共Fig. 2兲. It receives its time-dependent boundary conditions from HEC-HMS HEC-HMS is the updated version of the USACE rainfall-runoff model 共USACE-HEC 1998兲. It utilizes a graphical user interface to build a watershed model and to set up the precipitation and control variables for simulation. For this project, the Calaveras River watershed in California was modeled with a focus on the upper watershed that provides inflow into New Hogan Reservoir. A map of the location of the watershed is shown in Fig. 2. The watershed model created in HEC-HMS follows the form of the Sacramento District Corps office HEC-1 forecast model of the basin 共USACE 1987兲. This model utilizes one subbasin above New Hogan Reservoir that provides runoff into the reservoir. Below the reservoir, there is a routed channel reach and another subbasin that provides runoff to a sink point at Bellota. Bellota is a point on the lower Calaveras River where a flow and stage gauging station is located. A schematic description of the basin is shown in Fig. 3. 4. Calibration HEC-HMS Model Table Model parameter Fig. 3. HEC-HMS schematic description of Calaveras River basin Volume moisture deficit Wet Front Suction Conductivity % impervious Time of concentration Storage coefficient Recession constant Threshold flow Coefficients for Calaveras Watershed Calibrated value 0.1 in. 0.1 in. 0.125 in./h 22% 5.5 h 0.5 0.25 3,000 cfs 314 / JOURNAL OF HYDROLOGIC ENGINEERING / JULY/AUGUST 2002 Downloaded 28 Feb 2011 to 152.3.102.242. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org Fig. 4. Calibrated model versus observed New Hogan Reservoir inflows In Fig. 3, the outline of the watershed and branches of the upper Calaveras River are shown underneath the schematic components of the watershed model. The upper Calaveras watershed is represented by the subwatershed component Upper Basin. This subwatershed component provides inflows into New Hogan Reservoir, which is represented with a triangle. A second subwatershed is made up of the watershed below the reservoir and is titled Lower Basin. The gauging station at Bellota is represented in the watershed model as a sink, and the river reach between New Hogan Reservoir and Bellota is represented by a channel reach that routes streamflow using the Muskingum method. For this study, the inflows into New Hogan Reservoir from the upper watershed were the only components examined. In order to use the gridded precipitation component of HECHMS, a GIS-based spatial representation of the watershed is required. This spatial representation can be created using a set of UNIX-based ARC/INFO routines called GridParm 共USACE-HEC 1996兲. This representation treats the basin as a collection of cells that each have a response time based upon their relative location to the subwatershed outlet. Flow is routed from cell to cell based upon topography determined from the DEM using the ModClark method 共USACE 1996兲. One set of infiltration parameters is used for all the cells in the watershed. With this representation, HECHMS can transform spatial representations of rainfall data into runoff at the subwatershed outlet. Application Calibration and Verification of HEC-HMS The calibration period for the HEC-HMS model of the Calaveras basin was a 48-h period from February 8 to 9, 1999. Rainfall data were obtained from five rain gauges in the Calaveras Basin listed in Table 3. Calibration and verification were performed using rainfall data from these five ground-based rain gauges as point gauges. Final calibration parameters for the HEC-HMS model are shown in Table 4. The model uses a Green-Ampt infiltration/loss parameterization, the ModClark hydrograph transformation routine, and a recession base flow component. The initial loss and initial flow are treated as initial conditions and vary from simulation to simulation. A plot of the observed inflow into New Hogan Reservoir versus the model simulated inflow is shown in Fig. 4. Note that the observed and model simulated flows match well, except that the peak of the simulated flow is slightly later than the observed flow. The opposite occurs for the verification run. For verification purposes, a second 48-h period from February 16 to 18, 1999, was used. A plot of the observed versus predicted inflow into New Hogan Reservoir is shown in Fig. 5. Note that the predicted peak is early compared to the observed flow. The value in the ModClark hydrograph transformation routine was set to match the timing of the calibration and verification peaks as close as possible. Using this set of parameters for the Calaveras River watershed model, shown in Table 4, an application of the runoff forecasting process was conducted. Runoff Forecast Generation In order to demonstrate the use of Eta model forecast data to generate runoff forecasts with HEC-HMS, a 48-h forecast period from January 19 to 21, 1999, was selected. During this time period, rain gauges measured between 2 and 3 in. of rainfall in the upper part of the Calaveras basin. Fig. 6 shows the 48-h Eta forecasts of precipitation in the Calaveras River basin and its surrounding region for the above- Fig. 5. Verification results on predicted versus observed New Hogan Reservoir inflows JOURNAL OF HYDROLOGIC ENGINEERING / JULY/AUGUST 2002 / 315 Downloaded 28 Feb 2011 to 152.3.102.242. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org Fig. 6. Eta model forecast: 6-h accumulated precipitation over Calaveras River Basin from 1999-01-19 – 12:00Z to 1999-01-21 – 12:00Z Fig. 7. Sample of MM5 model forecasts of 1-h accumulated precipitation over Calaveras River basin mentioned time period of the study. There are eight plots in Fig. 6. Each plot represents the spatial distribution of the accumulated precipitation during a 6-h time period. Each square in the plots represents one AWIPS grid. These AWIPS grids have a size of roughly 40⫻40 km. In these plots, the heavy black outline indicates the boundary of the Calaveras River basin. The gray shades indicate the 6-h accumulated precipitation in mm in each grid. Six AWIPS grids cover the Calaveras River watershed, as shown in Fig. 6. Within each plot of Fig. 6, the number of grids is 4 in the vertical and 5 in the horizontal directions. The 48-h-ahead forecast of precipitation by the MM5 model covered the time period of 12:00 GMT January 19, 1999 to 12:00 GMT January 21, 1999. A sample of six hourly plots of the spatial distribution of the hourly precipitation depths generated by MM5 is shown in Fig. 7. In these plots, the heavy black outline indicates the boundary of the Calaveras River basin. The gray shades indicate the precipitation in mm/h in each grid. The spatial and temporal evolution of the precipitation fields in Calaveras River basin and its surrounding region are shown clearly in these plots. Using the precipitation inputs represented in Figs. 6 and 7, HEC-HMS was run with the gridded precipitation routine using the calibrated values shown in Table 4. For the Eta model forecast precipitation, the precipitation values were interpolated bilinearly in space and divided evenly in time in order to obtain hourly precipitation data on the HEC-HMS precipitation grid. A plot of observed inflows into New Hogan Reservoir compared with the HEC-HMS model using the Eta model precipitation, the MM5 model precipitation, and point gauge precipitation is shown in Fig. 8. Because of the interpolation in space and time on the Eta model forecast precipitation, the resulting precipitation values are too small to lead to any runoff using the point gauge calibrated model parameters of HEC-HMS. This can be seen in Fig. 8, where the Eta model runoff prediction is a recession curve only. Fig. 8. Comparison of observed runoff to HEC-HMS forecast runoff into New Hogan Reservoir using Eta model precipitation forecast, MM5 model precipitation forecast, and point gauge precipitation measurements 316 / JOURNAL OF HYDROLOGIC ENGINEERING / JULY/AUGUST 2002 Downloaded 28 Feb 2011 to 152.3.102.242. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org Fig. 9. Comparison of observed versus forecasted precipitation for three sites in Calaveras Basin Using the refined precipitation data from the MM5 simulation, the predicted runoff forecast is significantly less than the observed values, but does show a runoff response to the precipitation. Specifically, the magnitude of the forecast runoff is only 67% of the observed runoff and the forecast peak is 4 h later than the observed peak. The underprediction of the magnitude may be due in part to the fact that the calibration and verification were performed with point rain gauge data. Such a calibration may not be appropriate for gridded precipitation applications of HEC-HMS. In order to see if the forecast model could represent the appropriate runoff, the HEC-HMS model was run again with rainfall data taken from the five gauges used in the calibration process. The results are shown in Fig. 8. Note in Fig. 8 that, when the point precipitation gauge data is used with HEC-HMS, the flow is much better represented. The peak and timing of the peak are well simulated with the point gauge data. However, it must be noted that the results from the point rain gauge data can only be obtained after the rain events have been recorded by the rain gauges. This does not provide any lead time in terms of a forecast runoff. However, the MM5 results are forecast up to 48 h before the actual rain event occurs. As a further investigation into the differences between observed and forecast runoff by HEC-HMS, the differences between the MM5 forecast precipitation and point gauge precipitation values were compared. The point rainfall gauges used here are a subset of the rainfall gauges that were used to produce the model calibration and verification results of Figs. 4 and 5, and of forecast performance comparisons of Fig. 8. The rainfall hyetographs for the three gauges are compared with the nearest neighbor grid points within the MM5 simulation and are shown in Fig. 9. As can be seen from all of these plots, MM5 underpredicts the early part of the storm and overpredicts the later part of the storm. The MM5 total storm depths for all grid points fall in the range of 1– 4 in. with the nearest neighbor points to the gauges having similar depth totals between 2 and 3 in. The overall peaks of the rainfall intensities produced by MM5 also match those of the observed rainfall. The differences indicate that the calibration and verification process should include an evaluation of the different parameterizations in MM5 for representing precipitation processes. Such an evaluation would likely lead to a more refined precipitation forecast. Even without such an evaluation, the current results JOURNAL OF HYDROLOGIC ENGINEERING / JULY/AUGUST 2002 / 317 Downloaded 28 Feb 2011 to 152.3.102.242. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org point to a promising way of obtaining runoff predictions with a lead time much greater than is currently available. Conclusions Runoff forecasts from precipitation forecasts can be obtained using the methodology presented here. Eta model forecast products can be refined using the mesoscale atmospheric model MM5 to obtain hourly precipitation values over a 48-h period. The methodology presented here can obtain a runoff forecast within minutes of obtaining the precipitation forecast from the MM5 model. The process can be automated, yielding a valuable tool for reservoir management. The methodology was demonstrated for a 48-h forecast period in January 1999 in the Calaveras watershed in Northern California. Inflows to New Hogan Reservoir were predicted using Eta model forecast data and a refined precipitation forecast using the mesoscale model MM5. Because the Eta model precipitation forecast is given in 6-h intervals, the precipitation forecast must be distributed into an hourly format to be used in the current formulation of HEC-HMS. With the HEC-HMS model, which was calibrated by means of point gauge precipitation data, the timing and magnitude of the forecast peak in the runoff hydrograph were underestimated when the point gauge calibrated HEC-HMS model was driven by spatially distributed MM5 rainfall forecasts. However, when the point gauge calibrated HEC-HMS used point gauge rainfall for the same storm, the magnitude and timing of the peak runoff were matched. This would indicate that it is necessary to calibrate the HEC-HMS model with spatially distributed rainfall when using the model in the forecasting framework presented here. When the HEC-HMS model is calibrated and verified using the gridded precipitation data from MM5, runoff predictions will likely improve. The MM5 model can also be changed in terms of the choice of land surface model, boundary layer model, and local precipitation parameterization. These changes can be used to fine tune the precipitation forecast. The accuracy of the forecast for use in driving emergency management directives may not need to match the peak inflow exactly. The runoff forecast should, however, cross the appropriate thresholds so as to cause the appropriate directives to be enacted. Further work is required in order to provide quantitative measures for this type of forecast accuracy. Improved accuracy in terms of matching the timing and magnitude of the peak inflow and total volume of runoff would provide more information to reservoir operators for flood control releases. As a result, a new tool will be available for the prediction of runoff, which will provide an improved lead time for better reservoir operations as well as providing more lead time to act on emergency management directives. References Burks, J.E., and Staudenmaier, M.J. 共1996兲. ‘‘A comparison of the Eta and the Meso Eta models during the 11–12 December 1995 storm of the decade.’’ WR-Tech. Attachment 96-21. Charley, W. 共1986兲. ‘‘Weather radar as an aid to real-time water control.’’ MS thesis, Univ. of California, Davis, Calif. Coontz, R. 共1994兲. ‘‘Digital rivers—flood forecasting is about to get a badly needed overhaul.’’ Science, 34共4兲, 21. Dudhia, J. 共1993兲. ‘‘A nonhydrostatic version of the Penn State/NCAR mesoscale model: validation tests and simulation of an Atlantic cyclone and cold front.’’ Mon. Weather Rev., 121, 1493–1515. Franchini, M., and Lamberti, P. 共1994兲. ‘‘A flood routing Muskingum type simulation model based on level data alone.’’ Water Resour. Res., 30共7兲, 2183–2196. Grell, G., Dudhia, J., and Stauffer, D. 共1994兲. ‘‘A description of the fifthgeneration Penn State/NCAR mesoscale model 共MM5兲.’’ NCAR Tech. Note NCAR/TN-398⫹STR, National Center for Atmospheric Research, Boulder, Colo. James, W., Robinson, C., and Bell, J. 共1993兲. ‘‘Radar-assisted real-time flood forecasting.’’ J. Water Resour. Plan. Manage., 119共1兲, 32– 44. Janish, P.R., and Weiss, S.J. 共1996兲. ‘‘Evaluation of various mesoscale phenomena associated with severe convection during VORTEX-95 using the Meso Eta model.’’ Proc., 15th Conf. on Weather Analysis and Forecasting, American Meteorological Society, Boston, preprint. Kavvas, M., and Chen, Z. 共1998兲. ‘‘Meteorologic model interface for HEC-HMS NCEP Eta atmospheric model and HEC hydrologic modeling system.’’ Rep. prepared for the USACE-HEC, United States Army Corps of Engineers, Hydrologic Engineering Center, Davis, Calif. Lamberti, P., and Pilati, S. 共1996兲. ‘‘Flood propagation models for realtime forecasting.’’ J. Hydrol., 175共1-4兲, 239–265. Mimikou, M. 共1996兲. ‘‘Flood forecasting based on radar rainfall measurements.’’ J. Water Resour. Plan. Manage., 122共3兲, 151–156. National Weather Service. 共1997兲. ‘‘Disastrous floods from the severe winter storms in California, Nevada, Washington, Oregon, and Idaho: December 1996 –January 1997.’’ Natural Disaster Survey Rep., National Oceanographic and Atmospheric Administration, National Weather Service, Silver Spring, Md. Schneider, R.S., Junker, N.W., Eckert, M.T., and Considine, T.M. 共1996兲. ‘‘The performance of the 29 km Meso Eta model in support of forecasting at the Hydrometeorological Prediction Center.’’ Proc., 15th Conf. on Weather Analysis and Forecasting, American Meteorological Society, Boston, preprint. Staudenmaier, M.J. 共1996a兲. ‘‘A description of the Meso Eta model.’’ NWS Western Region Technical Attachment No. 96-06, NWS Western Region, Salt Lake City. Staudenmaier, M.J. 共1996b兲. ‘‘The initialization procedure in the Meso Eta model.’’ NWS Western Region Technical Attachment No. 96-30, NWS Western Region, Salt Lake City. U.S. Army Corps of Engineers 共USACE兲. 共1987兲. ‘‘Real-time flood forecasting and reservoir regulation in the Calaveras River Basin.’’ Project Rep. No. 87-2, Washington, D.C. U.S. Army Corps of Engineers Hydrologic Engineering Center 共USACEHEC兲. 共1996兲. GridParm procedures for deriving grid cell parameters for the ModClark rainfall-runoff model—user’s manual, USACE-HEC, Davis, Calif. U.S. Army Corps of Engineers Hydrologic Engineering Center 共USACEHEC兲. 共1998兲. HEC-HMS Hydrologic Modeling System user’s manual, USACE-HEC, Davis, Calif. Yapo, P., Sorooshian, S., and Gupta, V. 共1993兲. ‘‘A Markov chain flow model for flood forecasting.’’ Water Resour. Res., 29共7兲, 2427–2436. 318 / JOURNAL OF HYDROLOGIC ENGINEERING / JULY/AUGUST 2002 Downloaded 28 Feb 2011 to 152.3.102.242. Redistribution subject to ASCE license or copyright. Visithttp://www.ascelibrary.org