Application of linear regression methods based on model and experimental data to the retrieval of cloud liquid water path from ground-based microwave measurements
E.Yu. Biryukov, V.S. Kostsov
Saint Petersburg State University, 1, Ulyanovskaya str., Petrodvorets, St. Petersburg, 198504, Russia
Keywords: водозапас облаков, тропосфера, дистанционное зондирование, микроволновый радиометр, обратные задачи, регрессионное соотношение, cloud liquid water path, troposphere, remote sensing, microwave radiometer, inverse problems, linear regression
Abstract
The 14-channel RPG-HATPRO microwave radiometer has been operating at the Faculty of Physics, St. Petersburg State University since 2012. It performs continuous measurements of the cloud liquid water path (LWP). Along with the multiple quadratic regression (MQR) method provided by the instrument manufacturer, the “physical” retrieval algorithm, which is based on the inversion of the radiative transfer equation (IRTE), and the multiple linear regression (MLR) method are used. The estimates of the LWP retrieval errors are presented for the cases when MLR coefficients are derived from model calculations and from the experimental data (in the latter case, the IRTE results are used as reference data). It is shown that the application of the experimental data and the utilization of measurements in 7 spectral channels of the radiometer provide the LWP random error of 0.015-0.017 kg/m2 which is two times lower than in case of derivation of regression coefficients from model calculations. The bias does not exceed 0.005 kg/m2 in this case. It is demonstrated that the MLR results provide a reliable identification of clear-sky conditions if the criterion of minimal LWP variations is applied.
|
