ABSTRACT |
|
An adjoint method and 4-dimensional variational data assimilation (4DVAR)
are applied for a heavy rainfall case over the Korean Peninsula to access
the sensitivity of mesoscale convective systems to initial conditions, and
to obtain better initial data through data assimilation for mesoscale model
prediction. The active mesoscale convective systems responsible for the
heavy rainfall case evolved over the Yellow Sea where observed conventional
data are void.
Validity of linear approximation for the heavy rainfall case is investigated since the adjoint method is based upon the linear perturbation theory. The tangent linear model is valid up to 12 hours in describing the evolution of the initial random perturbations which are of comparable size to the magnitudes of analysis errors, although relatively larger errors occur over the heavy precipitation regions. Meanwhile, the validity of linear approximation is largely degraded by initial moisture perturbations. The sensitivity of evolution of the mesoscale convective systems to initial conditions is very localized, especially significant over the Yellow Sea where the convective system is initiated in the mid- and lower troposphere. Also, the largest adjoint sensitivity occurs in the moisture fields, then is followed by the temperature and wind fields. Under the favorable synoptic conditions for the heavy rainfall, slightly different initial conditions based on the intensity of initial disturbance and the vertical structure of moisture fields over the large sensible regions can yield much different evolution characteristics of the mesoscale convection and related precipitation. Initial perturbation, which plays an important role in the evolution of mesoscale convective system, is effectively analyzed by the technique of adjoint sensitivity. Moisture perturbation enhances moisture amounts in the mid- and lower troposphere, and thus provide a favorite condition for convective processes through the release of latent heat. Wind perturbation also intensifies the lower level convergence and relative vorticity. The 4DVAR experiments show that the adjustment of mass fields to wind fields in the mid- and lower troposphere is more effective. The assimilation of wind observations improves adjusted moisture fields in the assimilation window than the temperature assimilation does. Also, the assimilation of moisture observations produces approriate retrievals of the temperature and wind fields in the mid- and lower troposphere over the precipitation regions through the latent heat release of precipitation processes. Precipitation associated with a synoptic frontal system is not very sensitive to initial moisture fields and is reproduced by wind or/and temperature assimilation, whereas precipitation induced by the mesoscale convective system depends strongly upon the quality of initial moisture fields. Thus, the moisture assimilation can only reproduce reasonable precipitation pattern related to the mesoscale convective systems. Although the assimilation of precipitable water can improve the analysis of moisture structure, there is a large possibility of arbitrary adjustment of moisture fields since precipitable water contains vertically integrated moisture information. However, the assimilation using precipitable water observation along with wind and temperature observation reduces the possibility of the arbitrary adjustment, and produces better mesoscale precipitation pattern and amount of this case up to 24 hours. Therefore, the 4DVAR using remotely sensed precipitable water observation together with other observations could increase accurate moisture analysis and improve short-range precipitation forecasts for heavy rainfall. |