Using FSO Method to Evaluate SOUND Data Impacts on Forecast Error Reduction in North-West China

This study evaluates the impact of rawinsonde soundings (SOUND) observations on regional 24-hour forecast error reduction over the Urumqi regional numerical forecasting system domain using the adjoint-based forecast sensitivity to observation (FSO) system in WRFDA v3.8. The impact of SOUND observations on reducing the forecast error is assessed by conducting the WRFDA FSO continuous experiments from June 1 to September 30, 2015, including twice cold start time (0000UTC and 1200UTC) and once warm start time (0600UTC). As a note, the SOUND is artificially turned off at 1800 UTC. The impact is categorized for seven vertical levels: 1000, 925, 850, 700, 500, 300 and 100hPa. Based on dry energy, the results show that: (1) in regional numerical system, SOUND and SYNOP exhibit the largest observation impact on forecast error reduction, followed by AIREP. (2) Over all, SOUND assimilation presents the positive contribution at all vertical pressure levels at 0000UTC and 1200UTC. The impact is most positive at 0000UTC. However, the impact of SOUND assimilation is obviously negative at 0600UTC which is only positive at middle levels (700, 500 and 300hPa). (3) At 0000UTC, each kind of observational element is positive contribution on reducing forecast error at all vertical pressure levels. U, V and T at lower levels of atmosphere (925, 850 and 700hPa) and Q at the middle levels of atmosphere (500 and 300hPa) present a more positive contribution on forecast error reduction. Q is nearly all negative at 100hPa. At 1200UTC, the performance of elements assimilation on reducing forecast error is similar as 0000UTC except U is negative at 1000hPa and V is negative at 500hPa. At 0600UTC, T is almost negative contribution on forecast error reduction at all vertical pressure levels except 500hPa. U and Q are only positive impact on reducing forecast error at the middle levels of atmosphere (500 and 300hPa). V is positive from 850hPa to 300hPa. (4) The assimilated observation numbers of SOUND at 1000hPa and 925hPa are 25% and 50% of the numbers at higher levels because of the higher terrain elevation and time drift. The assimilated numbers of SOUND at 0600UTC are one third of the numbers at other start times. The encryption sounding data are concentrated in central and eastern China and most of them present negative impact on reducing forecast error, especially in T.