A Novel Bias Correction Scheme for the New York State Mesonet Microwave Radiometer Temperature Retrieval

Temperature retrievals from ground-based microwave radiometers (MWRs) are susceptible to systematic error (bias) arising from instrumental calibration issues and retrieval algorithm errors. In this study, we developed a novel bias correction (BC) scheme that effectively reduces the cold bias in the temperature retrievals from the MWR, deployed as a part of the New York State Mesonet (NYSM) Profiler Network for the last 8 years. In the training process, this BC scheme uses the High-Resolution Rapid Refresh analysis as reference to model the dependence between the observed brightness temperatures and the retrieved temperature bias using a multiple linear regression. In the application process, the BC scheme uses MWR observed brightness temperatures as the main predictors to predict the bias in the MWR temperature retrieval. Verifications using nearby radiosonde data from the National Weather Service and two other field campaigns demonstrate that the BC scheme effectively reduces cold biases throughout the lower troposphere and midtroposphere to less than 0.2 K and lowers the error standard deviation by approximately 10%-15%. In the presence of temperature inversion layers, previously identified as a major contributor to the cold bias, the BC scheme yields a smoother temperature increase near the inversion layer, significantly reducing temperature errors and bias at and above the inversion. In addition, the BC scheme can robustly correct the MWR temperature bias outside of the training window for at least 2-3 weeks into the future. This enables an ahead-of-time preparation of the BC scheme for real-time bias correction of the MWR temperature data.