Ground-Based Microwave Radiometer Variational Analysis during No-Rain and Rain Conditions

Abstract

Ground-based microwave radiometer (MWR) has been used for high-frequency retrievals of thermodynamic environments. However, raindrops on the radome of MWR and in the air cause errors in retrievals during precipitation events. Although a recent study has noted that off-zenith observations with neural networks (NN) reduce the retrieval errors, the effect of off-zenith observations with one-dimensional variational (1DVAR) technique, which is known to be more accurate than other methods, has not been studied. We developed a new 1DVAR technique that considers the effect of cloud liquid water. We statistically investigated the accuracy of vertical profiles of atmospheric temperature and water vapor retrieved by NN and 1DVAR techniques by using zenith and off-zenith observation at 15° elevation angle under no-rain and rainy conditions and compared them with results of radiosonde observations. The results showed that the 1DVAR technique outperforms NN and numerical model simulation in the estimation of thermodynamic profiles under no-rain conditions. The results also indicated that the error in retrieved profiles in the low-level troposphere can be reduced by the 1DVAR technique by using off-zenith observations even under rainy conditions with rainfall rate less than 1.0 mm h⁻¹, especially when the environment cannot be accurately reproduced by a numerical model.