The MP-3000A Microwave Radiometer is a passive remote sensing instrument with 35 vertically oriented bright-temperature observation channels capable of inverting atmospheric parameters based on atmospheric bright-temperature values. Microwave radiometer inversion accuracy can be affected by the atmospheric temperature background field. In order to select a suitable atmospheric temperature background field, this paper proposes a method to construct a physical model to generate the background. The physical model was set up according to the natural laws of atmospheric temperature. In order to verify its reasonableness, the atmospheric temperature profile generated from the background field is compared with the real atmospheric temperature profile. The results were analyzed by means of Frechette distance and root mean square error (RMSE). It shows that, compared with the fall and winter seasons, Atmospheric temperature background fields are more realistic in spring and summer. The root mean square error is 1.4664 and 1.4888. Compared to the height of 0-3km, the height of 3-7km and 7-10km, temperature profiles are more realistic in the middle and upper height. The errors are 1.6403 and 1.8894. The research in this paper is helpful for the accuracy improvement of atmospheric microwave remote sensing observation.
Microwave radiometer is a passive remote sensing instrument. The MP3000A microwave radiometer is used in this study. To address the problem of time-series augmentation of contour data, this paper uses a super-resolution network model based on convolutional neural network for time-series augm-entation of 34,568 temperature contour data measured by $M$ P3000A from April 22, 2023 to June 15, 2023 in Harbin. High- resolution branches are added to the original super-resolution network, and the results are compared and analyzed by normal-izing and non-normalizing the original data. The heatmap of the temperature data error shows that the time series amplified temperature data is weakly different from the true value, with a mean square error MSE of 2.4831 and a mean absolute error MAE of 1.3213, and the network model performs well. It is shown that the use of super-resolution network model works well in time series for atmospheric microwave remote sensing data under the influence of the Pacific monsoon climate.