The Use of Multi-Sensor Quantitative Precipitation Estimates for Deriving Extreme Precipitation Frequencies with Application in Louisiana

The Radar-based Quantitative Precipitation Estimates (QPE) is one of the NEXRAD products that are available in a high temporal and spatial resolution compared with gauges. Radar-based QPEs have been widely used in many hydrological and meteorological applications; however, a few studies have focused on using radar QPE products in deriving of Precipitation Frequency Estimates (PFE). Accurate and regionally specific information on PFE is critically needed for various water resources engineering planning and design purposes. This study focused first on examining the data quality of two main radar products, the near real-time Stage IV QPE product, and the post real-time RFC/MPE product. Assessment of the Stage IV product showed some alarming data artifacts that contaminate the identification of rainfall maxima. Based on the inter-comparison analysis of the two products, Stage IV and RFC/MPE, the latter was selected for the frequency analysis carried out throughout the study. The precipitation frequency analysis approach used in this study is based on fitting Generalized Extreme Value (GEV) distribution as a statistical model for the hydrologic extreme rainfall data that based on Annual Maximum Series (AMS) extracted from 11 years (2002-2012) over a domain covering Louisiana. The parameters of the GEV model are estimated using method of linear moments (L-moments). Two different approaches are suggested for estimating the precipitation frequencies; Pixel-Based approach, in which PFEs are estimated at each individual pixel and Region-Based approach in which a synthetic sample is generated at each pixel by using observations from surrounding pixels. The region-based technique outperforms the pixel based estimation when compared with results obtained by NOAA Atlas 14; however, the availability of only short record of observations and the underestimation of radar QPE for some extremes causes considerable reduction in precipitation frequencies in pixel-based and region-based approaches.