Impact of ENRR Observations on Global Weather Forecasts

Jih-Wang Aaron Wang

Tuesday, May 16, 2017, 2:00 pm
DSRC Room 2A305

Abstract

Three sets of 7-day 80-member ensemble forecasts, whose initial conditions differ by the inclusion or exclusion of the additional radiosonde and dropsonde observations of the 2015-2016 El Niño Rapid Response (ENRR) Field Campaign as well as by the data assimilation (DA) methods, have been made. By examining horizontal and vertical wind velocity, geopotential height, temperature, and precipitable water, the forecast root-mean-squared (RMS) errors with respect to the best analysis were found to grow with time. Adopting different DA methods to generate initial conditions induces significantly different forecast results in the 7-day course. The inclusion of the ENRR observations in the initial conditions help to signify El Niño features in the upper tropospheric wind and geopotential height fields via the advection of vorticity from the tropics to the extratropics. The inclusion also significantly reduces the forecast RMS errors in the first 24 hours in the tropics, and generates both improving and deteriorating results afterwards locally in the extratropics. Global mean error reduction for the upper tropospheric wind and geopotential height by ENRR observations is small.



The initial error reduction in the mid-tropospheric vertical velocity by the ENRR observations is concentrated in the seasonal strong precipitation area, while the initial reduction in the precipitable water is concentrated in the tropics especially over land. The error reduction in both fields quickly weakens in the first 24 hours. The GFS forecasts gradually develops dry and wet bias in precipitable water in several places, including wet bias in the tropical northwest Pacific Ocean and Central America, and dry bias in Indonesia and offshore Peru and Chile. The inclusion of the ENRR observations in the initial conditions induce 500hP vertical velocity forecast variants, which well correspond to precipitation forecast variants in the tropics. We conclude that the benefits of additional El Niño observations used in initial conditions are prominent but quickly decrease in the first 24 hours, while the forecast differences due to the different DA methods persist longer. The distinct initial conditions may also lead to upper tropospheric Rossby wave formation in the extratropics and different precipitation distribution in the tropics.