Bias Correction of Stage IV Multi-Sensor Precipitation Estimates in North Carolina

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 6 February 2014
Hall C3 (The Georgia World Congress Center )
Geneva M. Ely, North Carolina State University, Raleigh, NC; and A. M. Wootten and R. Boyles

Multi-sensor precipitation estimates (MPE) are a 4.765km x 4.765km gridded mosaic product created by National Centers for Environmental Prediction (NCEP) and National Weather Service (NWS) using radar-based precipitation data calibrated with surface rain gauges from an array of real-time hourly stations. MPE are used by local, state, and federal institutions for decision-making applications that require finer resolution precipitation totals. Previous research has shown MPE to have a root-mean-square error (RMSE) of 0.32 inches across the eastern United States when compared to observations from weather station networks independent of those used in the calculation of MPE (Wootten and Boyles, in review). This analysis aims to bias correct daily MPE in North Carolina to provide better quality estimates by using a higher density network of daily rain gauges from the NWS Cooperative Observer Network (COOP). A geographically weighted logistic regression model (GWLM) is utilized to determine the probability that the MPE grid cell value is greater than zero when the closest COOP stations to that grid cell reports non-zero precipitation values. These probabilities along with MPE bias are used to perform a Bayesian spatial quantile regression which produces bias corrected precipitation values. These values are then evaluated for accuracy against observed precipitation totals from the North Carolina Environment and Climate Observing Network (NC ECONet).