This COMET Cooperative Project combined proposals from two university researchers who were working with several operational forecasters from a Weather Forecast Office and a River Forecast Office. The first phase of the project, overseen by Dr. Ken Crawford, was designed to assist the NWS in making more skillful use of the WSR-88D through improved accuracy of estimated surface rainfall. Toward that end, the project involved validating the ability of the Oklahoma WSR-88D network to provide accurate estimates of surface rainfall under a variety of climatic extremes. The error characteristics of the WSR-88D were documented, the statistically relevant time and space scales of precipitation events determined, and the functional applicability of the Z/R relationship evaluated. Equally as important, the impact of precipitation errors on the accuracy of the hydrologic forecast was assessed.
The Oklahoma Mesonet, the Agricultural Research Service's micronetwork, and the WSR-88D network combine to represent a truly unique system that is capable of revealing the fine-scale spatial and temporal structure of surface rainfall patterns across Oklahoma. These resources have been accessed to compile a database of 50 cases that were used to:
1) Document the errors of the WSR-88D in estimating surface rainfall as a function of range from the radar, season, terrain, and climatic regime; convective versus stratiform; systems with low precipitation efficiency versus those with higImproved estimates of surface rainfall and river stage predictionsh efficiency; and mesoscale versus microscale. A total of 185 hours of WSR-88D precipitation data was calibrated using gauge observations from the Oklahoma Mesonetwork.
2) Determine optimum space/time combination of surface and radar-based estimates of rainfall
3) Evaluate the efficacy of the universally applied Z/R relationship
4) Recommend improvements for estimating surface rainfall where "improved estimates" are evaluated by their ability to reduce errors in river-stage forecasts when compared to those stage errors inherent in the real time operational forecasts
5) Summarize study results using a "decision tree" which was to be evaluated and validated through independent data sets obtained in 1995
The result of this research was a Ph.D. dissertation on the impact of calibrated radar estimates of rainfall on quantitative precipitation forecasts produced by extrapolation and by mesoscale modeling and the effect those estimates have on the stream-flow hydrograph produced by hydrologic modeling.
The second phase of the project, overseen by Dr. Baxter Vieux, was designed to improve knowledge of the hydrologic processes affecting flood prediction by using distributed rainfall data. Current flood modeling at the NWS River Forecast Centers used the Sacramento Model to forecast flood stages at gauge points. This model was used as a lumped watershed model that considered rainfall and other parameters affecting soil moisture to be constant for the entire basin. Incorporating spatially variable precipitation directly into the model is expected to yield better forecasts of the magnitude and timing of flood crests. Better soil moisture accounting can then be performed with distributed data, rather than by lumping spatially variable rainfall over large basins. With more accurate WSR-88D estimates (as obtained from the first phase of the project) and the use of a distributed model, it was expected that flash flood forecasting could be significantly improved.
Specific tasks in this part of the project involved:
1) Model development including: a) ingest of Stage III rainfall estimates, b) calibration of the model for storms that occur during the study period or selected historic storm events, c) validation for storms not used in the calibration, and d) integration as a module within the existing NWS river forecasting system.
2) Model testing for selected basins using rainfall excess data obtained from the Sacramento Model.
3) Incorporating digital terrain data for two test basins to ensure that the channel network and topography are accurately represented in the model. This would enable the NWS to forecast hydrographs at points interior to the basin, rather than only at gauging points as is the current practice. Testing was performed on historical and hypothetical storms to calibrate and validate hydraulic parameters.
The model development work was merged with the radar rainfall analysis work to compare the impact on hydrologic modeling of advanced processing with existing Stage III rainfall data for several rain events. The effects on forecast accuracy were evaluated using measured and simulated hydrograph.