Analysis of atmospheric temperature data by 4D spatial\u2013temporal statistical model

Ke Xu,Yaqiong Wang

doi:10.1038/s41598-021-98125-2

Abstract

The meteorological data such as temperature of the upper atmosphere is ssential for accurate weather forecasting. The Universal Rawinsonde Observation Program (RAOB) establishes an extensive radiosonde network worldwide to observe atmospheric meteorological data from the surface to the low stratosphere. The RAOB data data has very high accuracy but can offer a very limited spatial coverage. Meanwhile, ERA-Interim reanalysis data is widely available but with low-quality. We propose a 4D spatiotemporal statistical model which can make effective inferences from ERA-Interim reanalysis data to RAOB data. Finally, we can obtain a huge amount of RAOB data with high-quality and can offer a very wide spatial coverage. In empirical research, we collected data from 200 launch sites around the world in January 2015. The 4D spatiotemporal statistical model successfully analyzed the observation gaps at different pressure levels.

Highlights

Being able to accurately understand different temperatures at different latitudes, longitudes and altitudes is very important for research and practical applications[1,2,3,4]
We propose a 4D (3D space + 1D time) statistical model, it aims to estimate the observation gap between ERA-Interim reanalysis data and Rawinsonde Observation Program (RAOB) data
In “4D spatiotemporal statistical model” section, we present the 4D spatio-temporal statistical model

Summary

Model results

We first estimated the model by Expectation–Maximization algorithm. Second, we analyzed the model estimation results and verified the fitting goodness of our model based on cross-validation R2. In the atmosphere where the barometric pressure is lower than 400 hPa , the variance σǫ2(h) of the residuals increases, and βERA(h) deviates from the value 1. By comparing the predicted value with the real value for the selected 30% sites, we obtain the cross-validation R2 of the RAOB data model. The standard deviations of the spatial predictions at barometric pressures 450 hPa and[100] hPa , respectively. No matter what the altitude is, the variance of differences in areas where there are sampling sites are very small, while the variance of differences in areas far away from sampling sites are very large In this way, we explore the underlying mechanism of how the uncertainty changing at various barometric pressure levels

Conclusion

Additional information