902 An Investigation of the Forecast Performance for Precipitation in the Extended Global Ensemble Forecast System (GEFS) Reforecast

Wednesday, 10 January 2018
Exhibit Hall 3 (ACC) (Austin, Texas)
Eric Sinsky, NOAA/NCEP/EMC and IMSG, College Park, MD; and Y. Zhu, H. Guan, C. Melhauser, Y. Luo, and W. Li

Reforecasts are a useful tool for statistical post-processing since they can substantially improve forecast skill on the medium-range and subseasonal scales. In particular, information provided by reforecasts can be used to calibrate forecasts. Reforecasts are especially helpful for improving the skill of forecasts that contain extreme weather such as uncommon precipitation events. It is imperative to enhance forecast performance for precipitation events due to the effects precipitation can have on flooding. In order to ensure effective calibration with reforecast information, it is important to investigate the performance of the reforecast dataset.

The National Oceanic and Atmospheric Administration/ Environmental Modeling Center (NOAA/EMC) has created an ensemble reforecast dataset using the Global Ensemble Forecast System (GEFS) to support the Climate Prediction Center’s (CPC) operational mission and to satisfy the requirements of the SubX project. This reforecast for SubX is an 18-year (1999-2016) dataset initialized every Wednesday. The SubX reforecast, which has a different configuration as the operational GEFS v11, contains 11 members instead of 21 members and is integrated to 35 days instead of to 16 days. There are also updates to the stochastic physics, sea surface temperature (SST) and convection. According to an evaluation of a 2-year retrospective forecast, this configuration of the GEFS shows significantly improved forecast skill for 500 hPa geopotential height, 2-m temperature and precipitation compared to the operational GEFS v11.

The reforecast uses initial conditions from the Climate Forecast System Reanalysis (CFSR) from 1999 to 2010 and the operational GSI analysis from 2011 to 2016. The forecast precipitation variable that is to be evaluated is the 24-hr accumulated precipitation (APCP). APCP is verified against the Climatology-Calibration Precipitation Analysis (CCPA) over the Continental United States (CONUS). Some measures of verification that will be used to study precipitation prediction include mean error (bias), Frequency Bias (FBI) and the Equivalent Threat Score (ETS). The temporal and spatial structure of these precipitation verification measures will be examined for each reforecast year.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner