Retrieval of Atmospheric Profiles in the New York State Mesonet Using One‐Dimensional Variational Algorithm

Abstract

AbstractA one‐dimensional variational (1DVAR) algorithm was developed, which combines measurements taken by the ground‐based microwave radiometer profiler (MWRP) with the Rapid Refresh (RAP) model to retrieve a more accurate temperature and humidity profiles under clear sky. The passive and active microwave‐vector radiative transfer model was used to simulate brightness temperatures and calculate weighting functions for 22 channels of the MWRP. As MWRP measurements are mainly weighted in the lower atmosphere, a measurement adjustment is performed to reduce the contribution from above 10 km. The results of the 1DVAR algorithm have been compared with the MWRP built‐in neural network (NN) retrievals and radiosonde observations, which show that the 1DVAR method outperforms the NN retrieval both in temperature and water vapor. Our statistical study further shows that the water vapor profiles from RAP are biased higher than the radiosonde observations, while the 1DVAR‐retrieved humidity values show significant improvements throughout the entire 10‐km range. The improvements in temperature profiles occur mainly within the lowest 4‐km atmosphere, while upper level retrievals are mostly influenced by initial values. This is consistent with the characteristics of the Jacobian matrix and the background error covariance matrix. The outcomes of 44 clear‐sky cases show that the maximum mean retrieval error of the 1DVAR algorithm is less than 0.2 K for temperature below 4 km and less than 0.15 g/m3 in water vapor density. These results are shown to be significantly better than the NN retrieval (3.0 K and 1.25 g/m3) and also superior to the RAP reanalysis (0.3 K and 0.5 g/m3).