346172 An Evaluation of Quantitative Precipitation Estimators in the Western United States.

Tuesday, 26 June 2018
New Mexico/Santa Fe Room/Portal (La Fonda on the Plaza)
Matthew E. Jeglum, NOAA/NWS, Salt Lake City, UT; and P. Veals and C. Kahler

Accurate gridded Quantitative Precipitation Estimation (QPE) is critical as input to hydrologic models, to calibrating post-processed numerical weather model output, and situational awareness of operational weather forecasters. Due to the complexity of terrain in the Western US and the very tight gradients of precipitation that can occur there, creating high-quality QPE has historically been challenging in this region. Large areas of the Western US have minimal or no radar coverage from the WSR-88D network, and surface observations tend to be concentrated in valleys and populated areas, leaving much of the relatively wet high elevations unsampled. While the expanding Snowpack Telemetry (SNOTEL) network has helped to alleviate the latter problem, many areas are still unsampled. Integration of observations into the analysis, as well as extrapolation and interpolation of observations to unobserved areas, are critically-important part of any QPE product. We will assess how those methods perform at gauge locations in this paper.
There are currently a number of methods utilized in the United States to produce QPE, which vary from simple radar derived products to multi-sensor automated algorithms to heavily human-edited products. In this paper we will compare 6 different products computed at 6-h accumulation intervals over the Western US for a 12-month period from July 2016- July 2017. Analyzed QPE products include the radar-derived, gauge-corrected radar- derived and Mountain Mapper products from the Multi-radar/Multi-sensor System (MRMS), the QPE from the Unrestricted Mesoscale Analysis (URMA), QPE produced by the Parameter-elevation Regression on Independent Slopes Model (PRISM), and the Climatology-Calibrated Precipitation Analysis. Each product was compared to quality-controlled gauge data across the Western US to establish its quality using standard skill metrics, such as the Brier Skill Score.
The relative quality of each QPE product was found to vary by season and location within the Western US. URMA was the top performing QPE overall, while MRMS Mountain Mapper was the lowest performing. While all products are best in late winter and worst in late summer, this is particularly true of the MRMS Mountain Mapper. Despite that fact that many of the products utilize gauge data in creating the analysis, their performance at the gauge locations varied considerably. California and other coastal areas generally saw the best performance, while inland mountain regions displayed the lowest Brier Skill Scores. QPE biases tended to be low, especially in inland areas. PRISM displayed some of the best bias scores overall. Post hoc combinations of QPE products performed better than individual products in some cases.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner