Application Of 3.5-DVar Radar Data Assimilation Technic And Its Improved Scheme In WRF Model

Three-and-half-dimensional variational (3.5DVar) technic of radar dataassimilation and Physical initialization (PI) method are introduced detailly in thisarticle., and also the Doppler radar radial velocity and reflectivity data wereassimilated using this technic.3.5dVar, which is an expansion of the 3 DVar, using the consecutive volumescans, considering of time continuity, and by adding equations constrains, get thewind and the thermodynamic fields step by step. In this method, incrementalanalyses are performed in three steps to update the model state upon the backgroundstate, using three consecutive volume scans. First, radar radial-velocity observationsfrom three consecutive volume scans are analyzed on the model grid. The analyzedradial-velocity fields are then used in step-2 to estimate vector velocity fields at twotime levels between the three volume scans. The estimated vector velocity fields areused in step-3 to estimate thermodynamic fields at the central time level.Moreover, the action of weak constrains are discussed. Based on the discussion, anew improvement for 3.5Dvar is provided——retrieval vertical velocity fromreflectivity by PI method was added into the cost function of wind retrieval process,as an 'observation' information. The background for assimilation was provided byWRF model, so the interface problem is involved in this paper.A heavy rainfall process during 4th～5th July 2003 over the Huaihe River valleywas simulated by the WRF Model. Comparing the wind and thermodynamic fromthe 3.5Dvar and PI3.5Dvar, the result shows the difference is associated with thehorizontal distribution of reflectivity observations. There are positive increment of verticalvelocity at echoes region, but there is no obvious response in the horizontal velocityfield. The result of equation constrain test also shows that the adjustment of windfield by equation constrains is very small, which might be the result of inconsistencebetween the simplified equation and the small scale weather system. Finally,comparision of the 1～2hr precipitation prediction shows that the assimilation of radardata has prominently improved the forecasting of rainfall region, and theprecipitation predicted by PI3.5Dvar is bigger than the old one.