Microwave Radiometer Remote Sensing Atmospheric Temperature And Humidity Profile In Yangjiang

With the emergence of the foundation microwave radiometer, not only the ordinary atmospheric sounding data can be compensated, to overcome the limitation of the conventional atmospheric sounding in the application, but also can obtain new and valuable meteorological data in the study of atmospheric structure. It is one of the important methods of atmospheric sounding.This article has the following several points:1. This paper uses the microwave radiometer observation data to retrieval the atmospheric temperature profiles and the water vapor density profiles with the multiple linear regression method and the BP neural network method. The multiple linear regression method is still one of the main and effective methods, which can extract as much information as possible from the priori data, take the correlation of the ground measurement into account, add the corresponding ground measurements into the linear regression equation to improve the accuracy of the retrieval profiles. The BP neural network method has a strong advantage in dealing with nonlinear problems, theoretically, it can approximate any nonlinear problems instead of specially designed complex inversion algorithm, and it can save a lot of trouble to the direct analysis of the physical model because of independent on the physical forward model.2. This paper analyzes the two methods of retrieval accuracy. To the atmospheric temperature profiles regression, for the multiple linear regression method, the deviation is not greater than4K below the height of6.5km, the deviation increases with increasing altitude, and the whole deviation is not greater than6K; for the BP neural network method, the deviation is not greater than IK below the height of8.5km, the deviation increases with increasing altitude, and the whole deviation is not greater than2K. To the water vapor density profiles regression, for the multiple linear regression method, the deviation increases with increasing altitude, after about1.5km reached a maximum, about4.5km they began to decrease, and the whole deviation is not greater than4g/m3; for the BP neural network method, the trends is the same as the multiple linear regression method, and the whole deviation is not greater than2g/m3. The result shows that: the inversion result of BP neural network method is closer to radiosonde data than that of multiple linear regression method.3. This paper also have the inversion of liquid water path and precipitable water vapor, and analysis the results with micro-pulse lidar observations, in order to know the correspondence between rainfall, cloud thickness, cloud height and liquid water path, precipitable water vapor. The rainfall and cloud thickness have large impact on it. From the change of the two parameters with time, the use of microwave remote sensing can not only remote sensing of liquid water path and precipitable water vapor, but can also continuously monitor the weather, cloud and rain change, and atmospheric water vapor fluctuations and changes in cloud liquid water, it can play a role in the short-term weather forecast application.This paper has innovative research in the following areas:conventional radiosonde data does not contain the brightness temperature, so we need to simulate the brightness temperature of three-year history sounding. Yangjiang is located on the beach, it is cloudy, so the estimate of cloud water will become an important and difficult work. This paper reference a variety of literature and the western pacific tropical water cloud amount and cloud shape statistics, add the local cloud liquid water content into the model, without distinguishing the sunny samples and sky samples.