Microwave radiance data assimilation (DA) enhances initial conditions for numerical weather prediction (NWP) and shows great potential for improving forecasts in tropical regions like East Africa, where observational data scarcity and complex tropical dynamics present significant challenges. Effectiveness of radiance assimilation is a function of variations in channel sensitivity to local atmospheric conditions and region-specific bias characteristics. However, microwave radiance assimilation in Limited-Area Models (LAMs) over East Africa remains largely unexplored. This study investigates the impact of assimilating microwave radiance channels with weighting functions peaking in the troposphere and lower stratosphere on rainfall forecasts over East Africa from a five-satellite constellation: the Microwave Temperature Sounder-2 (MWTS-2) onboard Fengyun-3D (FY-3D), the Advanced Technology Microwave Sounder (ATMS) onboard JPSS, and the Advanced Microwave Sounding Unit-A (AMSU-A) onboard NOAA-15/18/19 satellites. The 6-h cycling DA experiments over a convectively active 15-day period show that assimilation of ATMS and AMSU-A radiances enhances representation of initial conditions, thereby reducing analysis and forecast errors. Assimilation of MWTS-2 radiances improves the analysis and forecasts further, especially for the tropospheric thermodynamic fields. The joint multi-microwave assimilation fills critical observation gaps over East Africa, allowing realistic simulations of diurnal precipitation trends, and capturing rainfall intensities exceeding 50 mm in 24 h, especially for T+12-h to T+24-h lead times. These findings are validated by a high-intensity rainfall case over Mandera, where spatio–temporal consistency is observed in instability and convection triggering. Forecast evaluation metrics have confirmed enhanced rainfall forecast skill for deep and rapidly developing convective systems. The study provides valuable insights into the gains of assimilating microwave radiance data over tropical regions, particularly in East Africa.

For numerical weather prediction (NWP), data assimilation (DA) combines short-term forecasts and various atmospheric observations to achieve optimal initial conditions, based on which subsequent forecasts are launched. With the rapid advancements in numerical models and observing systems, DA has been significantly evolved. Modern methods now can account for uncertainties of state variables across various spatiotemporal scales, incorporate multiscale observation error statistics, and enforce dynamical constrains and model balances. Meanwhile, observations from various platforms, such as ground-based, aircraft, and satellite, have been assimilated. These include data from polar-orbiting and geostationary satellites, radar-derived radial winds and reflectivity, Global Navigation Satellite System (GNSS) radio occultations, etc. To further utilize the advanced observing systems and DA techniques for high-impact weather predictions, target observation strategies have been developed to identify areas where additional observations can yield the greatest predict improvements. Based on the advancements of DA theories and methods, China’s operational systems have made significant progress, establishing advanced operational DA systems. Over the past decade, the forecast skill of 5-day global weather prediction has improved by approximately 15%. The article reviews a century of development in DA, and discusses future directions, including the advanced DA methods, operational frameworks, integration of novel observations, and the synergy between DA and artificial intelligence.

The traditional variational assimilation method uses isotropic and homogeneous background error covariance, which ignores the weather system dependence of the background error covariance, and the introduction of ensemble flow-dependent background error covariance in the variational framework requires additional ensemble forecasts. In order to introduce more reasonable background error covariance in the variational assimilation, a "cloud-dependent" background error covariance is constructed by introducing cloud indices, and a cloud-dependent background error covariance assimilation scheme is proposed and applied to the assimilation of radar and other multi-source observations. Based on the cloud-dependent background error covariance data assimilation scheme, a series of single observation tests and batch cyclic assimilation forecasts during the rainy season as well as detailed diagnostic analysis of rainfall cases are carried out. From the single observation tests, it is found that the cloud-dependent background error covariance can dynamically adjust the background error at each grid point in real time, resulting in significant anisotropy and dependence of the analysis increments on cloud and rain characteristics; the batch cyclic assimilation and forecasting experiments show that the radar assimilation with cloud-dependent background error covariance can steadily improve the precipitation forecasting capability, and the improvement is especially obvious for the large magnitude precipitation scores; The diagnosis of strong convective storms further shows that the application of cloud-dependent background error covariance improves the prediction of dynamical, thermal, water vapor and hydrometeor fields. The assimilation scheme based on cloud-dependent background error covariance can introduce background error covariance information that is more consistent with real-time weather characteristics in the framework of variational assimilation, which provides a basis for better assimilation of high-resolution radar data and effectively improves rainfall forecasting.

Abundant mesoscale hydrometeor information is included in dual polarimetric radar data, which plays an essential role in the monitoring of convective storms. To assimilate dual polarimetric radar data more effectively, a direct dual polarimetric radar data assimilation scheme is developed. The tangent linear and adjoint examination for observation operators and its adjoint, single observation tests, and cycling data assimilation and forecasting experiments in a real-data case are conducted. The results in tangent linear and adjoint examination indicate that this dual polarimetric radar data direct assimilation scheme satisfies the requirement of precision, and the operator is reasonably constructed. It can be seen from the results of single observation tests that the observation information from dual polarimetric radars can be transferred to relative variables such as hydrometeors via hydrometeor background error covariance, and the analysis becomes more cooperated. The results from cycling assimilation and forecasting experiments in a real-data case show that dynamic, thermodynamic and microphysical structures can be optimized by dual polarimetric radar data assimilation, and the capability of rainfall prediction is improved. It can be concluded that the direct dual polarimetric radar data assimilation scheme under variational framework can make the dual polarimetric radar observation information to be assimilated more reasonably, and the quality of 6-hour prediction can be improved. It has more computing efficiency compared with ensemble Kalman filter, and is more convenient for operational application.

Assimilation of the initial aerosol field can improve the accuracy of coupled meteorology-aerosol forecasts in WRF-Chem. To discuss the influence of aerosol assimilation on coupled model forecasts at different times for a heavy haze pollution process in October 2015, an experimental study of joint meteorology-aerosol assimilation and short-term forecast at four different times of a day was conducted. The results show that assimilating aerosol observations on the basis of assimilated meteorological data can improve the overestimation of simulation caused by the emission source inventory with large uncertainty, and reduce the positive bias of PM_2.5 in the initial field. With decreased PM_2.5 concentration in the initial field, the reduction of surface radiation caused by aerosols is weakened, which increases downward shortwave radiation in the 6 h forecast in the daytime. The response of near-surface temperature and humidity to radiation changes (warming and humidity reduction) during the daytime is spatially coincident over a large area and temporally synchronized, and continues even at night under the influence of circular rolling forecast. The structure of the near-surface warming and humidity reduction contributes to upward development of the boundary layer, which in turn promotes upward transport of aerosols and ultimately attenuates the overestimation of surface PM_2.5 in the forecast. Consequently, the improvement of the initial PM_2.5 information due to aerosol assimilation makes the 6 h coupled forecasts more accurate.

To investigate the impact of radial velocity assimilation in the clear-air region on forecasting, this paper conducted cycle assimilation experiments for short-duration convection and heavy rainfall forecasting using the latest classified raw radar data from the Meteorological Observation Centre of China Meteorological Administration. These experiments involve the assimilation of radial velocity in the clear-air region, alongside the assimilation of radar reflectivity factor and radial velocity in the precipitation region. The results show that as the cycle assimilation progresses, radial velocity in the clear-air region gradually influences the interior of the convective system, adjusting the thermal and dynamic conditions within the convection. This adjustment improves the coordination between various physical quantities, thereby improving the forecast skill of short-duration convection. Furthermore, the radial velocity in the clear-air region gradually affects the physical quantities in the precipitation region through the cycle of assimilation, thereby improving precipitation forecast skills. The study discloses that reasonable assimilation of radial velocity in the clear-air region can play an important role in pre-convection and precipitation forecasts.