The present study designs a data assimilation framework for all-sky GMI radiances in the Weather Research Forecast (WRF) model for the prediction of tropical cyclone events over the Bay of Bengal (BOB) region. Experiments are performed using the RTTOV-SCATT radiative transfer model for three cyclone events (Hudhud (2014), Vardah (2016), and Kyant (2016)) for 19 V, 23 V, 37 V, 89 V, 166 V, 183±3, and 183±7 V channels. Results show that observed and simulated brightness temperatures (Tb) agree well for low-frequency channels (<89 GHz). However, significant discrepancies do exist when simulating low Tb at high frequency channels. Therefore, a set of 26 simulations is considered using snow particle shapes from single scattering property databases. A new methodology is devised in order to compare the 26 simulations with observations; taking into account the uncertainties on hydrometeor radiative properties. Results reveal that model simulations are characterized by an underestimation of occurrences within several Tb ranges (~200–300 K). The probability distribution function (PDF) of normalized observed-minus first guess is close to a Gaussian shape after applying a dedicated model of observation errors. The potential to integrate the GMI sensor data within a WRF data assimilation system is thus demonstrated.

本研究设计了气象研究预报（WRF）模型中所有天空GMI辐射的数据同化框架，以预测孟加拉湾（BOB）地区上空的热带气旋事件。使用RTTOV-SCATT辐射传递模型针对19 V，23 V，37 V，37 V，89 V，166 V，183±3的三个气旋事件（Hudhud（2014），Vardah（2016）和Kyant（2016））进行了实验3和183±7 V通道。结果表明，对于低频通道（<89 GHz），观察到的和模拟的亮度温度（Tb）吻合得很好。但是，在高频信道上模拟低Tb时，确实存在显着差异。因此，使用来自单个散射特性数据库的雪粒形状考虑了一组26个模拟。设计了一种新的方法，以便将26个模拟与观察结果进行比较。考虑到水凝磁辐射特性的不确定性。结果表明，模型模拟的特征是低估了几个Tb范围（〜200–300 K）内的事件。在应用专用的观察误差模型后，标准化的观察到的负第一次猜测的概率分布函数（PDF）接近高斯形状。因此，展示了将GMI传感器数据集成到WRF数据同化系统中的潜力。