Rainfall Assimilation through an Optimal Control of Initial and Boundary Conditions in a Limited-Area Mesoscale Model

Abstract

To assess the impact of rainfall observations on short-range forecasts of precipitation, and to improve our understanding of the physical processes responsible for the development of a mesoscale convective system (MCS) associated with the dryline that occurred on 10 April 1979 in the midwestern United States, a series of four-dimensional variational data assimilation experiments was conducted based on the special dataset collected in the Severe Environmental Storm and Mesoscale Experiment. A nonhydrostatic mesoscale model (MM5) with a relatively simple moist physics and its adjoint were used for both the model simulation and data assimilation. A previous numerical simulation of this MCS, based on conventional initialization procedures, failed to correctly simulate the location and intensity of the observed rainfall. This is attributed to the lack of mesoscale details in the model's initial conditions for the low-level moisture convergence and the upper-level disturbances related to the upper-level jet streak. In contrast, the initial conditions created by the four-dimensional variational data assimilation method, which incorporated 3-h rainfall data along with wind, temperature, surface moisture, and precipitable water measurements, produced an improved short-range (up to 12 h) rainfall prediction. It also captured many important mesoscale features including the structure of MCSs, the lower- and upper-level jets, the position of the dryline, the low-level moisture convergence, and the formation of a localized capping inversion (lid). In addition, the spinup time required for precipitation was reduced. Additional experiments were conducted to assess the importance of lateral boundary conditions (LBCs) in the assimilation procedure, the importance of the precipitable water measurements, and the impact of moist physics. In comparison to the experiment in which only initial conditions (ICs) are used as a control variable, controlling both the initial and lateral boundary conditions during the minimization procedure produced a closer match to the observed rainfall while fewer changes are made to the analyzed ICs. The authors showed that assimilating precipitable water into the model is very important. The precipitable water assimilation constrains the large-scale model moisture error growth while allowing the model to generate mesoscale structures through rainfall assimilation. The 4DVAR rainfall assimilation experiments using two different cumulus parameterization schemes produced very similar adjustments to the original analysis, and model forecasts from the ?optimal? ICs and LBCs obtained through rainfall assimilation using a cumulus parameterization scheme different from the one used in the 4DVAR procedure were seen to perform better than that from CTRL without 4DVAR. These results strongly confirm that improved quantitative precipitation forecasts of mesoscale convective systems are possible through the assimilation of rainfall observations, along with other conventional data. Further improvement can be expected with the use of a high-resolution model with improved moist physics and boundary layer parameterization.